Big data and machine learning algorithms for health-care delivery. Lancet Oncol. 2019;20:e262-e273. [PMID: 31044724 DOI: 10.1016/s1470-2045(19)30149-4]

Efficient and Effective Diabetes Care in the Era of Digitalization and Hypercompetitive Research Culture: A Focused Review in the Western Pacific Region with Malaysia as a Case Study

There are approximately 220 million (about 12% regional prevalence) adults living with diabetes mellitus (DM) with its related complications, and morbidity knowingly or unconsciously in the Western Pacific Region (WP). The estimated healthcare cost in the WP and Malaysia was 240 billion USD and 1.0 billion USD in 2021 and 2017, respectively, with unmeasurable suffering and loss of health quality and economic productivity. This urgently calls for nothing less than concerted and preventive efforts from all stakeholders to invest in transforming healthcare professionals and reforming the healthcare system that prioritizes primary medical care setting, empowering allied health professionals, improvising health organization for the healthcare providers, improving health facilities and non-medical support for the people with DM. This article alludes to challenges in optimal diabetes care and proposes evidence-based initiatives over a 5-year period in a detailed roadmap to bring about dynamic and efficient healthcare services that are effective in managing people with DM using Malaysia as a case study for reference of other countries with similar backgrounds and issues. This includes a scanning on the landscape of clinical research in DM, dimensions and spectrum of research misconducts, possible common biases along the whole research process, key preventive strategies, implementation and limitations toward high-quality research. Lastly, digital medicine and how artificial intelligence could contribute to diabetes care and open science practices in research are also discussed.

Risk prediction model for chemotherapy-induced nausea and vomiting in cancer patients: a systematic review

BACKGROUND

Chemotherapy-induced nausea and vomiting increase the healthcare burden and lead to adverse clinical outcomes in cancer patients. Although many risk prediction models for chemotherapy-induced nausea and vomiting have been developed, their methodological quality and applicability remain uncertain.

OBJECTIVES

To systematically review and evaluate existing studies on risk prediction models for chemotherapy-induced nausea and vomiting in cancer patients.

METHODS

PubMed, the Cochrane Library, Embase, Web of science, CINAHL, Scopus, China National Knowledge Infrastructure (CNKI), China Science and Technology Journal Database (VIP), Wanfang Database, Chinese Biomedical literature Database (CBM) were systematically searched from inception to October 1, 2024. Studies were appraised critically and data extracted by two authors independently based on the Prediction Model Risk of Bias Assessment Tool (PROBAST) and Data Extraction for Systematic Reviews of Prediction Modeling Studies (CHARMS).

RESULTS

A total of 4195 articles were retrieved, ultimately including 17 studies with 62 models for chemotherapy-induced nausea and vomiting. The sample size of the included studies ranged from 137 to 2215, with areas under the curve ranging from 0.602 to 0.850. In this study, the deep forest model demonstrated strong discrimination and calibration, outperforming conventional machine learning and traditional regression models. The five most important predictors in the deep forest model were creatinine clearance, age, sex, anticipatory nausea and vomiting, and antiemetic regimen. Across all included studies, age, chemotherapy regimens, cycles of chemotherapy, history of alcohol consumption, prior episodes of chemotherapy-induced nausea and vomiting, sleep quality before chemotherapy, sex, antiemetic regimens, history of morning sickness, anticipatory nausea and vomiting, were the most frequently reported predictors. All studies were rated as high risk of bias mainly due to poor reporting of the participants and analysis domains, with high concerns regarding applicability in 9 studies.

CONCLUSION

The research on prediction models for chemotherapy-induced nausea and vomiting model is in its developing stage, with both commonalities and differences in predictors. Despite the overall acceptable performance of chemotherapy-induced nausea and vomiting models, most studies have methodological shortcomings, and few models have been validated. Future studies should refer to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) guideline for model design, implementation, and reporting. Moreover, studies with larger sample sizes and multicenter external validation are necessary to enhance the robustness of predictive models.

REGISTRATION

The protocol for this study is registered with PROSPERO (registration number: CRD42024505012).

Comparing logistic regression and machine learning for obesity risk prediction: A systematic review and meta-analysis

BACKGROUND

Logistic regression (LR) has traditionally been the standard method used for predicting binary health outcomes; however, machine learning (ML) methods are increasingly popular.

OBJECTIVE

This study aimed to compare the performance of ML and LR for obesity risk prediction, identify how LR and ML were being compared, and identify the commonly used ML methods.

METHODS

We conducted comprehensive searches in PubMed, Scopus, Embase, IEEE Xplore, and Web of Science databases on 24th November 2023, with no restrictions on publication dates. Meta-analyses were performed to quantify the overall predictive performance of the methods using the area under the curve (AUC) for LR, AUC for the best performing ML, as well as the difference in the AUC between the two approaches as the effect measures.

RESULTS

We included 28 studies out of 913 abstracts screened. Accuracy and sensitivity were the most commonly used performance measures. More than half of the studies used AUC, with no calibration assessment conducted in any of the studies. Decision trees followed by boosting algorithms were the most commonly used ML methods. Seventy-five percent of the studies were at high risk of bias. There were 14 included studies in the meta-analysis. The pooled AUC for LR was 0.75 (95% CI 0.70 to 0.80) and the pooled AUC for ML was 0.76 (95% CI 0.70 to 0.82). The pooled difference in logit(AUC) between ML and LR was 0.13 (95% CI -0.11 to 0.37).

CONCLUSION

We conclude that there is no significant difference in the performance of ML and LR for obesity risk prediction. However, there is a need for improved quality of reporting of studies, the use of more performance measures particularly calibration, and to validate models in different populations.

Proactive vs. passive algorithmic ethics practices in healthcare: the moderating role of healthcare engagement type in patients' responses

BACKGROUND

Artificial intelligence (AI) is transforming healthcare, but concerns about algorithmic biases and ethical challenges hinder patient acceptance. This study examined the effects of proactive versus passive algorithmic ethics practices on patient responses across different healthcare engagement types (privacy-focused vs. utility-focused).

METHODS

We conducted a 2 × 2 online experiment with 513 participants in China. The experiment manipulated the healthcare provider's algorithmic ethics approach (proactive vs. passive) and the healthcare engagement type (privacy-focused vs. utility-focused). Participants were randomly assigned to view a scenario describing a hospital's AI diagnostic system, then completed measures of attitudes, trust, and intentions to use the AI-enabled service.

RESULTS

Proactive algorithmic ethics practices significantly increased positive attitudes, trust, and usage intentions compared to passive practices. The positive impact of proactive practices was stronger for privacy-focused healthcare (e.g., mental health services) compared to utility-focused services emphasizing care optimization.

CONCLUSIONS

This study underscores the critical role of proactive, context-specific algorithmic ethics practices in cultivating patient trust and engagement with AI-enabled healthcare. To optimize outcomes, healthcare providers must strategically adapt their ethical governance approaches to align with the unique privacy-utility considerations that are most salient to patients across different healthcare contexts and AI use cases.

CLINICAL TRIAL NUMBER

Not applicable.

Preoperative blood and CT-image nutritional indicators in short-term outcomes and machine learning survival framework of intrahepatic cholangiocarcinoma

BACKGROUND&AIMS

Intrahepatic cholangiocarcinoma (iCCA) is aggressive with limited treatment and poor prognosis. Preoperative nutritional status assessment is crucial for predicting outcomes in patients. This study aimed to compare the predictive capabilities of preoperative blood like albumin-bilirubin (ALBI), controlling nutritional status (CONUT), prognostic nutritional index (PNI) and CT-imaging nutritional indicators like skeletal muscle index (SMI), visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT), visceral to subcutaneous adipose tissue ratio (VSR) in iCCA patients undergoing curative hepatectomy.

METHODS

290 iCCA patients from two centers were studied. Preoperative blood and CT-imaging nutritional indicators were evaluated. Short-term outcomes like complications, early recurrence (ER) and very early recurrence (VER), and overall survival (OS) as long-term outcome were assessed. Six machine learning (ML) models, including Gradient Boosting (GB) survival analysis, were developed to predict OS.

RESULTS

Preoperative blood nutritional indicators significantly associated with postoperative complications. CT-imaging nutritional indicators show insignificant associations with short-term outcomes. All preoperative nutritional indicators were not effective in predicting early tumor recurrence. For long-term outcomes, ALBI, CONUT, PNI, SMI, and VSR were significantly associated with OS. Six ML survival models demonstrated strong and stable performance. GB model showed the best predictive performance (C-index: 0.755 in training cohorts, 0.714 in validation cohorts). Time-dependent ROC, calibration, and decision curve analysis confirmed its clinical value.

CONCLUSION

Preoperative ALBI, CONUT, and PNI scores significantly correlated with complications but not ER. Four Image Nutritional Indicators were ineffective in evaluating short-term outcomes. Six ML models were developed based on nutritional and clinicopathological variables to predict iCCA prognosis.

The integration of artificial intelligence into clinical medicine: Trends, challenges, and future directions

BACKGROUND AND OBJECTIVES

AI has emerged as a transformative force in clinical medicine, changing the diagnosis, treatment, and management of patients. Tools have been derived for working with ML, DL, and NLP algorithms to analyze large complex medical datasets with unprecedented accuracy and speed, thereby improving diagnostic precision, treatment personalization, and patient care outcomes. For example, CNNs have dramatically improved the accuracy of medical imaging diagnoses, and NLP algorithms have greatly helped extract insights from unstructured data, including EHRs. However, there are still numerous challenges that face AI integration into clinical workflows, including data privacy, algorithmic bias, ethical dilemmas, and problems with the interpretability of "black-box" AI models. These barriers have thus far prevented the widespread application of AI in health care, and its possible trends, obstacles, and future implications are necessary to be systematically explored. The purpose of this paper is, therefore, to assess the current trends in AI applications in clinical medicine, identify those obstacles that are hindering adoption, and identify possible future directions. This research hopes to synthesize evidence from other peer-reviewed articles to provide a more comprehensive understanding of the role that AI plays to advance clinical practices, improve patient outcomes, or enhance decision-making.

METHODS

A systematic review was done according to the PRISMA guidelines to explore the integration of Artificial Intelligence in clinical medicine, including trends, challenges, and future directions. PubMed, Cochrane Library, Web of Science, and Scopus databases were searched for peer-reviewed articles from 2014 to 2024 with keywords such as "Artificial Intelligence in Medicine," "AI in Clinical Practice," "Machine Learning in Healthcare," and "Ethical Implications of AI in Medicine." Studies focusing on AI application in diagnostics, treatment planning, and patient care reporting measurable clinical outcomes were included. Non-clinical AI applications and articles published before 2014 were excluded. Selected studies were screened for relevance, and then their quality was critically appraised to synthesize data reliably and rigorously.

RESULTS

This systematic review includes the findings of 8 studies that pointed out the transformational role of AI in clinical medicine. AI tools, such as CNNs, had diagnostic accuracy more than the traditional methods, particularly in radiology and pathology. Predictive models efficiently supported risk stratification, early disease detection, and personalized medicine. Despite these improvements, significant hurdles, including data privacy, algorithmic bias, and resistance from clinicians regarding the "black-box" nature of AI, had yet to be surmounted. XAI has emerged as an attractive solution that offers the promise to enhance interpretability and trust. As a whole, AI appeared promising in enhancing diagnostics, treatment personalization, and clinical workflows by dealing with systemic inefficiencies.

CONCLUSION

The transformation potential of AI in clinical medicine can transform diagnostics, treatment strategies, and efficiency. Overcoming obstacles such as concerns about data privacy, the danger of algorithmic bias, and difficulties with interpretability may pave the way for broader use and facilitate improvement in patient outcomes while transforming clinical workflows to bring sustainability into healthcare delivery.

Mucoepidermoid carcinoma: Enhancing diagnostic accuracy and treatment strategy through machine learning models and web-based prognostic tool

BACKGROUND

Oral cancer, particularly mucoepidermoid carcinoma (MEC), presents diagnostic challenges due to its histological diversity and rarity. This study aimed to develop machine learning (ML) models to predict survival outcomes for MEC patients and pioneer a clinically accessible prognostic tool.

METHODS

Using the SEER database (2000-2020), we constructed predictive models with five ML algorithms: Random Forest Classifier (RFC), Gradient Boosting Classifier (GBC), Logistic Regression (LR), K-Nearest Neighbors (KNN), and Multilayer Perceptron (MLP). Predictive variables were identified via Cox regression, and Kaplan-Meier analysis assessed survival trends. Model performance was validated through the area under the curve (AUC) of receiver operating characteristic (ROC) curves.

RESULTS

This study included 1314 patients diagnosed with MEC of the oral cavity. The RFC demonstrated the highest predictive accuracy (AUC = 0.55), followed by the GBC and RFC (AUC = 0.53). The most affected primary site was the hard palate, followed by the retromolar and cheek mucosa. Survival rates varied with the treatment modality, with the highest rates observed in patients undergoing surgery alone. ML models have identified age, sex, and metastasis as significant prognostic factors influencing survival outcomes, underscoring the complexity and heterogeneity of MEC.

CONCLUSIONS

This study highlights ML's potential to enhance survival predictions and personalize treatment for MEC patients. We developed the first web-based prognostic tool, providing a novel, accessible solution for improving clinical decision-making in MEC.

An Appraisal of the Quality of Development and Reporting of Predictive Models in Spine Surgery

Study DesignLiterature review.ObjectiveThe Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) statement was developed to improve the generalizability of predictive models. This study systematically evaluated the quality of predictive models related to spine procedures and assessed their compliance with the TRIPOD guidelines.MethodsA systematic search was conducted on PubMed to identify original research articles published between January 1st, 2018, and February 1st, 2023 reporting prediction models in the top six spine journals ranked by Scimago Journal Ranking (SJR): Journal of Bone and Joint Surgery, Spine, Journal of Orthopaedic Trauma, Journal of Neurosurgery: Spine, Neurosurgery, and Neurosurgical Focus. We assessed article adherence to the TRIPOD criteria using a standardized checklist.Results72 articles were included and analyzed with the TRIPOD checklist. Median compliance with the TRIPOD criteria was 57.14% (IQR: 48.33-64.95%). Compliance varied significantly across journals (P < 0.05). Among the TRIPOD criteria, the lowest compliance was observed in blinding the assessment of predictors (n = 8, 16.00%), fully presenting the model for use (n = 12, 17.91%), and providing sufficient information to allow for the external validation of results (n = 13, 19.70%).ConclusionsPublished machine learning models predicting outcomes in spine surgery often do not meet the established guidelines for their development, validation, and reporting outlined by TRIPOD. This lack of compliance may suggest that these models have not been adequately validated externally or adopted into routine clinical practice in spine surgery.

Integrating Radiogenomics and Machine Learning in Musculoskeletal Oncology Care

Musculoskeletal tumors present a diagnostic challenge due to their rarity, histological diversity, and overlapping imaging features. Accurate characterization is essential for effective treatment planning and prognosis, yet current diagnostic workflows rely heavily on invasive biopsy and subjective radiologic interpretation. This review explores the evolving role of radiogenomics and machine learning in improving diagnostic accuracy for bone and soft tissue tumors. We examine integrating quantitative imaging features from MRI, CT, and PET with genomic and transcriptomic data to enable non-invasive tumor profiling. AI-powered platforms employing convolutional neural networks (CNNs) and radiomic texture analysis show promising results in tumor grading, subtype differentiation (e.g., Osteosarcoma vs. Ewing sarcoma), and predicting mutation signatures (e.g., TP53, RB1). Moreover, we highlight the use of liquid biopsy and circulating tumor DNA (ctDNA) as emerging diagnostic biomarkers, coupled with point-of-care molecular assays, to enable early and accurate detection in low-resource settings. The review concludes by discussing translational barriers, including data harmonization, regulatory challenges, and the need for multi-institutional datasets to validate AI-based diagnostic frameworks. This article synthesizes current advancements and provides a forward-looking view of precision diagnostics in musculoskeletal oncology.

Multi-view contrastive learning and symptom extraction insights for medical report generation

The task of generating medical reports automatically is of paramount importance in modern healthcare, offering a substantial reduction in the workload of radiologists and accelerating the processes of clinical diagnosis and treatment. Current challenges include handling limited sample sizes and interpreting intricate multi-modal and multi-view medical data. In order to improve the accuracy and efficiency for radiologists, we conducted this investigation. This study aims to present a novel methodology for medical report generation that leverages Multi-View Contrastive Learning (MVCL) applied to MRI data, combined with a Symptom Consultant (SC) for extracting medical insights, to improve the quality and efficiency of automated medical report generation. We introduce an advanced MVCL framework that maximizes the potential of multi-view MRI data to enhance visual feature extraction. Alongside, the SC component is employed to distill critical medical insights from symptom descriptions. These components are integrated within a transformer decoder architecture, which is then applied to the Deep Wrist dataset for model training and evaluation. Our experimental analysis on the Deep Wrist dataset reveals that our proposed integration of MVCL and SC significantly outperforms the baseline model in terms of accuracy and relevance of the generated medical reports. The results indicate that our approach is particularly effective in capturing and utilizing the complex information inherent in multi-modal and multi-view medical datasets. The combination of MVCL and SC constitutes a powerful approach to medical report generation, addressing the existing challenges in the field. The demonstrated superiority of our model over traditional methods holds promise for substantial improvements in clinical diagnosis and automated report generation, indicating a significant stride forward in medical technology.

Serum calcium-based interpretable machine learning model for predicting anastomotic leakage after rectal cancer resection: A multi-center study

BACKGROUND

Despite the promising prospects of utilizing artificial intelligence and machine learning (ML) for comprehensive disease analysis, few models constructed have been applied in clinical practice due to their complexity and the lack of reasonable explanations. In contrast to previous studies with small sample sizes and limited model interpretability, we developed a transparent eXtreme Gradient Boosting (XGBoost)-based model supported by multi-center data, using patients' basic information and clinical indicators to forecast the occurrence of anastomotic leakage (AL) after rectal cancer resection surgery. The model demonstrated robust predictive performance and identified clinically relevant thresholds, which may assist physicians in optimizing perioperative management.

AIM

To develop an interpretable ML model for accurately predicting the occurrence probability of AL after rectal cancer resection and define our clinical alert values for serum calcium ions.

METHODS

Patients who underwent anterior resection of the rectum for rectal carcinoma at the Department of Digestive Surgery, Xijing Hospital of Digestive Diseases, Air Force Medical University, and Shaanxi Provincial People's Hospital, were retrospectively collected from January 2011 to December 2021,. Ten ML models were integrated to analyze the data and develop the predictive models. Receiver operating characteristic (ROC) curves, calibration curve, decision curve analysis, accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and F1 score were used to evaluate model performance. We employed the SHapley Additive exPlanations (SHAP) algorithm to explain the feature importance of the optimal model.

RESULTS

A total of ten features were integrated to construct the predictive model and identify the optimal model. XGBoost was considered the best-performing model with an area under the ROC curve (AUC) of 0.984 (95%confidence interval: 0.972-0.996) in the test set (accuracy: 0.925; sensitivity: 0.92; specificity: 0.927). Furthermore, the model achieved an AUC of 0.703 in external validation. The interpretable SHAP algorithm revealed that the serum calcium ion level was the crucial factor influencing the predictions of the model.

CONCLUSION

A superior predictive model, leveraging clinical data, has been crafted by employing the most effective XGBoost from a selection of ten algorithms. This model, by predicting the occurrence of AL in patients after rectal cancer resection, has identified the significant role of serum calcium ion levels, providing guidance for clinical practice. The integration of SHAP provides a clear interpretation of the model's predictions.

A quantitative analysis of the use of anonymization in biomedical research

Anonymized biomedical data sharing faces several challenges. This systematic review analyzes 1084 PubMed-indexed studies (2018-2022) using anonymized biomedical data to quantify usage trends across geographic, regulatory, and cultural regions to identify effective approaches and inform implementation agendas. We identified a significant yearly increase in such studies with a slope of 2.16 articles per 100,000 when normalized against the total number of PubMed-indexed articles (p = 0.021). Most studies used data from the US, UK, and Australia (78.2%). This trend remained when normalized by country-specific research output. Cross-border sharing was rare (10.5% of studies). We identified twelve common data sources, primarily in the US (seven) and UK (three), including commercial (seven) and public entities (five). The prevalence of anonymization in the US, UK, and Australia suggests their practices could guide broader adoption. Rare cross-border anonymized data sharing and differences between countries with comparable regulations underscore the need for global standards.

Machine learning in ovarian cancer: a bibliometric and visual analysis from 2004 to 2024

OBJECTIVE

Ovarian cancer (OC) is a common malignant tumor in women, with poor prognosis and high mortality rates. Early diagnosis, screening, and prognostic prediction of OC have long been focal points and challenges in this field. In recent years, machine learning (ML) has gradually demonstrated its unique advantages in the early diagnosis, screening, and prognostic prediction of tumors, including OC.This study aims to analyze global development trends and research hotspots in the application of ML for OC, thereby providing a reference for future research directions.

METHODS

We searched the Web of Science Core Collection (WoSCC) for all publications related to OC and ML from 2004 to 2024, conducting a quantitative analysis using VOSviewer, R software, and CiteSpace.

RESULTS

A total of 777 articles were retrieved.The number of publications related to ML and OC has grown continuously over the past 20 years.China led with 254 articles.The most prominent journals include Gynecologic Oncology, Nature, Clinical Cancer Research, Cancer Research, and Journal of Clinical Oncology.Research hotspots are: (a) ML-driven OC biomarker discovery and personalized treatment; (b) ML in tumor microenvironment analysis and resistance prediction; (c) ML in imaging-based diagnosis and risk stratification; (d) ML in multicenter OC studies.

CONCLUSION

ML in OC is currently in a developmental phase and shows promising potential for the future. This study provides researchers and clinicians with a more systematic understanding of research priorities and forthcoming developments in this area.

A comprehensive analysis identified an autophagy-related risk model for predicting recurrence and immunotherapy response in stage I lung adenocarcinoma

Background

Lung adenocarcinoma (LUAD) is characterized by early recurrence and poor prognosis. Autophagy is a double-edged sword in tumor development and anti-tumor therapy resistance. However, the prediction of relapse and therapeutic response in LUAD patients with stage I based on the signature of autophagy remains unclear.

Methods

Gene expression data were obtained from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) database. Autophagy-associated genes were extracted from the Human Autophagy Moderator Database. The autophagy score was established by Least Absolute Shrinkage and Selection Operator (LASSO) regression. Real-time PCR was used to detect gene expression of hub genes in LUAD patients. Protein-protein interaction (PPI) was analyzed to identify crucial genes. Gene set enrichment analysis (GSEA) was used to reveal the molecular features of patients. ESTIMATE algorithm was applied to estimate the tumor immune infiltration. TIDE score and Genomics of Drug Sensitivity in Cancer (GDSC) database were used to assess therapeutic response.

Results

We established an autophagy score based on 19 autophagy genes. Among these genes, MAP1LC3B played a crucial role in PPI network and was down-regulated in tumor tissues both in TCGA and local cohort. Receiver operating characteristic (ROC) curve showed that the risk model effectively predict RFS of stage I LUAD (area under the curve (AUC) at 1, 2, 3 years = 0.701, 0.836, and 0.818, respectively). Multivariate regression analysis indicated that the autophagy score was an independent predictor for relapse (P < 0.001, HR = 4.8, 95% CI [3.25-7.2]). The autophagy score also showed great predictive efficacy in the external validation GEO cohorts. GSEA revealed gene sets significantly enriched in immunity, cell cycle, and adhesion, etc. Meanwhile, we found the autophagy score was negatively related to KRAS mutation (P = 0.017) but positively associated with TP53 mutation (P = 6.4e-11). The autophagy score had a negative relationship with CD8+, CD4+ T cell, and dendritic cell, and positively correlated with immune checkpoint molecule CD276. Patients with a high autophagy score were sensitive to chemotherapy and targeted therapy, while resistant to immune checkpoint inhibitors.

Conclusion

We constructed an effective recurrence risk predictive model for stage I LUAD patients based on autophagy related genes. High autophagy score predicted a higher recurrence risk and suppressing tumor immune microenvironment.

Functions, interactions and prognostic role of POLE: a bioinformatics analysis

OBJECTIVE

To describe POLE characteristics and reported mutations in endometrial cancer (EC) and analyze the impact of these mutations on the structure and function of the protein, as well as their relationship with the survival and prognosis of the disease.

METHODS

We retrieved reported mutations for POLE in EC from Catalogue of Somatic Mutations in Cancer database. We analyzed the most frequent mutations possible impact in the protein using HOPE server. We built a protein-protein network using Network Analyst, Cytoscape, and Network Analyzer plugin for topological analysis, enrichment analysis was performed using Gene Ontology: Biological processes. Clinical data was retrieved from cBioPortal database to compare overall survival between mutated POLE (POLEmut) and wild-type POLE. Relation of mutational status of POLE in EC and immune cell infiltration was analyzed using CIBERSORT algorithm in TIMER2.0 server.

RESULTS

Thirty mutations in POLE were retrieved, most reported mutations were p.P286R, p.V411L and p.A456P, these mutations were likely to be pathogenic. Network analysis of POLE showed interaction of this protein in biological processes such as DNA repair, the cell proliferation cycle, and mechanisms of resistance to platinum. Immune infiltration analysis showed that T cell CD8+, T cell memory activated CD4+, T cell follicular helper, T cell gamma delta and macrophage M1 were more infiltrated in EC POLEmut tumors.

CONCLUSION

Mutations in POLE might affect DNA polymerase epsilon function. These mutations also affect interactions with other proteins like proteins involved in different DNA repairing mechanisms. POLE mutations may lead to platinum resistance, but they can also trigger an immune response that improves prognosis.

Predicting lack of clinical improvement following varicose vein ablation using machine learning

OBJECTIVE

Varicose vein ablation is generally indicated in patients with active/healed venous ulcers. However, patient selection for intervention in individuals without venous ulcers is less clear. Tools that predict lack of clinical improvement (LCI) after vein ablation may help guide clinical decision-making but remain limited. We developed machine learning (ML) algorithms that predict 1-year LCI after varicose vein ablation.

METHODS

The Vascular Quality Initiative database was used to identify patients who underwent endovenous or surgical varicose vein treatment for Clinical-Etiological-Anatomical-Pathophysiological (CEAP) C2 to C4 disease between 2014 and 2024. We identified 226 predictive features (111 preoperative [demographic/clinical], 100 intraoperative [procedural], and 15 postoperative [immediate postoperative course/complications]). The primary outcome was 1-year LCI, defined as a preoperative Venous Clinical Severity Score (VCSS) minus postoperative VCSS of ≤0, indicating no clinical improvement after vein ablation. The data were divided into training (70%) and test (30%) sets. Six ML models were trained using preoperative features with 10-fold cross-validation (Extreme Gradient Boosting [XGBoost], random forest, Naïve Bayes classifier, support vector machine, artificial neural network, and logistic regression). The primary model evaluation metric was area under the receiver operating characteristic curve (AUROC). The algorithm with the best performance was further trained using intraoperative and postoperative features. The focus was on preoperative features, whereas intraoperative and postoperative features were of secondary importance, because preoperative predictions offer the most potential to mitigate risk, such as deciding whether to proceed with intervention. Model calibration was assessed using calibration plots, and the accuracy of probabilistic predictions was evaluated with Brier scores. Performance was evaluated across subgroups based on age, sex, race, ethnicity, rurality, median Area Deprivation Index, prior ipsilateral varicose vein ablation, location of primary vein treated, and treatment type.

RESULTS

Overall, 33,924 patients underwent varicose vein treatment (30,602 endovenous [90.2%] and 3322 surgical [9.8%]) during the study period and 5619 (16.6%) experienced 1-year LCI. Patients who developed the primary outcome were older, more likely to be socioeconomically disadvantaged, and less likely to use compression therapy routinely. They also had less severe disease as characterized by lower preoperative VCSS, Varicose Vein Symptom Questionnaire scores, and CEAP classifications. The best preoperative prediction model was XGBoost, achieving an AUROC of 0.94 (95% confidence interval [CI], 0.93-0.95). In comparison, logistic regression had an AUROC of 0.71 (95% CI, 0.70-0.73). The XGBoost model had marginally improved performance at the intraoperative and postoperative stages, both achieving an AUROC of 0.97 (95% CI, 0.96-0.98). Calibration plots showed good agreement between predicted and observed event probabilities with Brier scores of 0.12 (preoperative), 0.11 (intraoperative), and 0.10 (postoperative). Of the top 10 predictors, 7 were preoperative features including VCSS, Varicose Vein Symptom Questionnaire score, CEAP classification, prior varicose vein ablation, thrombus in the greater saphenous vein, and reflux in the deep veins. Model performance remained robust across all subgroups.

CONCLUSIONS

We developed ML models that can accurately predict outcomes after endovenous and surgical varicose vein treatment for CEAP C2 to C4 disease, performing better than logistic regression. These algorithms have potential for important utility in guiding patient counseling and perioperative risk mitigation strategies to prevent LCI after varicose vein ablation.

Optimizing Stroke Risk Prediction: A Primary Dataset-Driven Ensemble Classifier With Explainable Artificial Intelligence

Background and Aims

Stroke remains a leading cause of mortality and long-term disability worldwide, presenting a significant global health challenge. Effective early prediction models are essential for reducing its impact. This study introduces a novel ensemble method for predicting stroke using two datasets: a primary dataset collected from a hospital, containing medical histories and clinical parameters, and a secondary dataset.

Methods

We applied several preprocessing techniques, including outlier detection, data normalization, k-means clustering, and missing value detection, to refine the datasets. A novel ensemble classifier was developed, combining AdaBoost, Gradient Boosting Machine (GBM), Multilayer Perceptron (MLP), and Random Forest (RF) algorithms to enhance predictive accuracy. Additionally, Explainable Artificial Intelligence (XAI) techniques such as SHAP and LIME were integrated to elucidate key features influencing stroke prediction.

Results

The proposed ensemble classifier achieved an accuracy of 95% for the secondary dataset and 80.36% for the primary dataset. Comparative analysis with other machine learning models highlighted the superior performance of the ensemble approach. The integration of XAI further provided insights into the critical indicators influencing stroke classification, improving model interpretability and decision-making.

Conclusion

Our study demonstrates that the novel ensemble classifier, supported by effective preprocessing and XAI techniques, is a powerful tool for stroke prediction. The high accuracy rates achieved validate its effectiveness and potential for practical clinical application. Future work will focus on incorporating deep learning techniques and medical imaging to further improve classification accuracy and model performance.

Machine learning prediction models for multidrug-resistant organism infections in ICU ventilator-associated pneumonia patients: Analysis using the MIMIC-IV database

OBJECTIVE

This study aims to construct and compare four machine learning models using the MIMIC-IV database to identify high-risk factors for multidrug-resistant organism (MDRO) infection in Ventilator-associated pneumonia (VAP) patients.

METHODS

The study included 972 VAP patients from the MIMIC-IV database. Data encompassing demographic information, vital signs, laboratory results, and other relevant variables were collected. The class imbalance issue was addressed using the Synthetic Minority Over-sampling Technique (SMOTE). The dataset was randomly split into training and testing sets (8:2). LASSO regression and feature importance scores were used for feature selection. Clinical prediction models were built using logistic regression, XGBoost, random forest and gradient boosting machine. The performance of the models was evaluated through receiver operating characteristic(ROC) curve analysis.Model calibration was assessed using calibration curves and Brier scores. The effectiveness was evaluated through Decision Curve Analysis (DCA). SHAP was utilized for model interpretation.

RESULTS

Among 972 patients, 824 were non-MDROs-VAP and 128 were MDROs-VAP. Comparative analysis revealed statistically significant differences in various clinical parameters. XGBoost exhibited the best predictive performance, incorporating 20 features with an AUC of 0.831 (95 % CI: 0.785-0.877) on the test set. Calibration curves demonstrated robust consistency, corroborated by Decision Curve Analysis (DCA) affirming the clinical utility. SHAP analysis identified the most important features: red cell distribution width, duration of mechanical ventilation, anion gap, basophil percentage, and neutrophil percentage.

CONCLUSION

This study established and compared four machine learning models for MDROs infections in VAP patients. XGBoost was identified as the optimal predictor, and SHAP values provided insights into 20 independent risk factors, confirming its excellent predictive value.

IMPLICATIONS FOR CLINICAL PRACTICE

VAP is a common infection in ICU patients with a heightened risk of MDRO and increased mortality. The recognition of high bias in existing models calls for future research to employ rigorous methodologies and robust data sources, aiming to develop and validate more accurate and clinically applicable predictive models for MDROs infections in VAP patients.

Adoption challenges to artificial intelligence literacy in public healthcare: an evidence based study in Saudi Arabia

In recent years, Artificial Intelligence (AI) is transforming healthcare systems globally and improved the efficiency of its delivery. Countries like Saudi Arabia are facing unique adoption challenges in their public healthcare, these challenges are specific to AI literacy, understanding and effective usage of AI technologies. In addition, cultural, regulatory and operational barriers increase the complication of integrating AI literacy into public healthcare operations. In spite of its critical contribution in enabling sustainable healthcare development, limited studies have addressed these adoption challenges. Our study explores the AI literacy adoption barriers in context to Saudi Arabian public healthcare sector, focusing on its relevance for advancing healthcare operations and achieving Sustainable Development Goals (SDGs). The research aims to identifying and addressing the adoption challenges of Artificial Intelligence literacy within the public healthcare in Saudi Arabia. The research aims to enhance the understanding of AI literacy, its necessity for enhancing healthcare operations, and the specific hurdles that impede its successful AI adoption in Saudi Arabia's public healthcare ecosystem. The research employs a qualitative analysis using the T-O-E framework to explore the adoption challenges of AI literacy. Additionally, the Best-Worse Method (BWM) is applied to evaluate the adoption challenges to AI literacy adoption across various operational levels within Saudi Arabia's public healthcare supply chain. The study uncovers substantial adoption challenges at operational, tactical, and strategic level, including institutional readiness, data privacy, and compliance with regulatory frameworks. These challenges complicate the adoption of AI literacy in the Saudi public healthcare supply chains. The research offers critical insights into the various issues affecting the promotion of AI literacy in Saudi Arabia's public healthcare sector. This evidence-based study provides essential commendations for healthcare professionals and policymakers to effectively address the identified challenges, nurturing an environment beneficial to the integration of AI literacy and advancing the goals of sustainable healthcare development.

Artificial intelligence tool development: what clinicians need to know?

Digital medicine and smart healthcare will not be realised without the cognizant participation of clinicians. Artificial intelligence (AI) today primarily involves computers or machines designed to simulate aspects of human intelligence using mathematically designed neural networks, although early AI systems relied on a variety of non-neural network techniques. With the increased complexity of the neural layers, deep machine learning (ML) can self-learn and augment many human tasks that require decision-making on the basis of multiple sources of data. Clinicians are important stakeholders in the use of AI and ML tools. The review questions are as follows: What is the typical process of AI tool development in the full cycle? What are the important concepts and technical aspects of each step? This review synthesises a targeted literature review and reports and summarises online structured materials to present a succinct explanation of the whole development process of AI tools. The development of AI tools in healthcare involves a series of cyclical processes: (1) identifying clinical problems suitable for AI solutions, (2) forming project teams or collaborating with experts, (3) organising and curating relevant data, (4) establishing robust physical and virtual infrastructure, and computer systems' architecture that support subsequent stages, (5) exploring AI neural networks on open access platforms before making a new decision, (6) validating AI/ML models, (7) registration, (8) clinical deployment and continuous performance monitoring and (9) improving the AI ecosystem ensures its adaptability to evolving clinical needs. A sound understanding of this would help clinicians appreciate the development of AI tools and engage in codesigning, evaluating and monitoring the tools. This would facilitate broader use and closer regulation of AI/ML tools in healthcare settings.

Exploring the mechanisms of Chaige Kangyi Recipe in treating recurrent pregnancy loss with insulin resistance

Chinese herbal medicine effectively treats recurrent pregnancy loss, though its mechanism is unclear. This study used RStudio 4.3.0 to collect successful cases for cluster analysis, identifying medication patterns and core formulas, and further researching key prescriptions. The prescription is frequently used for recurrent abortion patients with insulin resistance. UPLC-QTOF-MS identified components, and network pharmacology explored key prescription targets in recurrent abortion with insulin resistance, validated by molecular docking and in vitro experiments. Traditional Chinese medicine treatment for 177 recurrent abortion cases and 640 prescriptions was analysed using RStudio 4.3.0 to identify medication patterns. Chaige Kangyi Recipe (CGKYR) active components and targets were obtained from TCMNPAS, and a herb-ingredient-target gene network was constructed using Cytoscape 3.7.2. GeneCards provided RSA target genes, and Cytoscape visualised a drug-disease target PPI network. Metascape software performed GO and KEGG enrichment analyses. UHPLC-MS/MS identified active compounds in core prescriptions, and molecular docking evaluated the therapeutic effects and mechanisms of major chemical components on key targets. Key prescriptions were derived from RStudio 4.3.0 cluster analysis of the Chaige Kangyi Recipe (CGKYR), commonly used for recurrent miscarriages with insulin resistance. Sixty-seven active ingredients were identified via UPLC-QTOF-MS. Network pharmacology revealed 179 target genes related to CGKYR's effects on recurrent miscarriage with insulin resistance. PPI analysis indicated IL-6, AKT1, STAT3, and INS as potential targets. Molecular docking demonstrated strong binding activity of four compounds with IL-6.CCK-8 assays showed CGKYR promoted HDSC proliferation dose-dependently. In vitro experiments indicated CGKYR increased IL-6 mRNA expression in human decidual stromal cells. CGKYR employs a multifaceted therapy for RPL complicated by insulin resistance, enhancing endometrial receptivity and stimulating HDSC proliferation by upregulating IL-6 mRNA expression in human decidual stromal cells.

A Hybrid Model-Based Clinicopathological Features and Radiomics Based on Conventional MRI for Predicting Lymph Node Metastasis and DFS in Cervical Cancer

This study aimed to improve the accuracy of the diagnosis of lymph node metastasis (LNM) and prediction of patient prognosis in cervical cancer patients using a hybrid model based on MRI and clinical aspects. We retrospectively analyzed routine MR data from 485 patients with pathologically confirmed cervical cancer from January 2014 to June 2021. The data were divided into a training cohort (N = 261), internal cohort (N = 113), and external validation cohort (n = 111). A total of 2194 features were extracted from each ROI from T2WI and CE-T1WI. The clinical model (M1) was built with clinicopathological features including squamous cell carcinoma antigen, MRI-reported LNM, maximal tumor diameter (MTD). The radiomics model (M2) was built with four radiomics features. The hybrid model (M3) was constructed with squamous cell carcinoma antigen, MRI-reported LNM, MTD which consists of M1 and four radiomics features which consist of M2. GBDT algorithms were used to create the scores of M1 (clinical-score, C-score), M2 (radiomic score, R-score), and M3 (hybrid-score, H-score). M3 showed good performance in the training cohort (AUCs, M3 vs. M1 vs. M2, 0.917 vs. 0.830 vs. 0.788), internal validation cohorts (AUCs, M3 vs. M1 vs. M2, 0.872 vs. 0.750 vs. 0.739), and external validation cohort (AUCs, M3 vs. M1 vs. M2, 0.907 vs. 0.811 vs. 0.785). In addition, higher scores were significantly associated with worse disease-free survival (DFS) in the training cohort and the internal validation cohort (C-score, P = 0.001; R-score, P = 0.002; H-score, P = 0.006). Radiomics models can accurately predict LNM status in patients with cervical cancer. The hybrid model, which incorporates clinical and radiomics features, is a novel way to enhance diagnostic performance and predict the prognosis of cervical cancer.

Estimated glucose disposal rate outperforms other insulin resistance surrogates in predicting incident cardiovascular diseases in cardiovascular-kidney-metabolic syndrome stages 0-3 and the development of a machine learning prediction model: a nationwide prospective cohort study

BACKGROUND

The American Heart Association recently introduced the concept of cardiovascular-kidney-metabolic (CKM) syndrome, highlighting the increasing importance of the complex interplay between metabolic, renal, and cardiovascular diseases (CVD). While substantial evidence supports a correlation between the estimated glucose disposal rate (eGDR) and CVD events, its predictive value compared with other insulin resistance (IR) indices, such as triglyceride-glucose (TyG) index, TyG-waist circumference, TyG-body mass index, TyG-waist-to-height ratio, triglyceride-to-high density lipoprotein cholesterol ratio, and the metabolic score for insulin resistance, remains unclear.

METHODS

This prospective cohort study utilized data from the China Health and Retirement Longitudinal Study (CHARLS). The individuals were categorized into four subgroups based on the quartiles of eGDR. The associations between eGDR and incident CVD were evaluated using multivariate logistic regression analyses and restricted cubic spline. Seven machine learning models were utilized to assess the predictive value of the eGDR index for CVD events. To assess the model's performance, we applied receiver operating characteristic (ROC) and precision-recall (PR) curves, calibration curves, and decision curve analysis.

RESULTS

A total of 4,950 participants (mean age: 73.46 ± 9.93 years), including 50.4% females, were enrolled in the study. During follow-up between 2011 and 2018, 697 (14.1%) participants developed CVD, including 486 (9.8%) with heart disease and 263 (5.3%) with stroke. The eGDR index outperformed six other IR indices in predicting CVD events, demonstrating a significant and linear relationship with all outcomes. Each 1-unit increase in eGDR was associated with a 14%, 14%, and 19% lower risk of CVD, heart disease, and stroke, respectively, in the fully adjusted model. The incorporation of the eGDR index into predictive models significantly improved prediction performance for CVD events, with the area under the ROC and PR curves equal to or exceeding 0.90 in both the training and testing sets.

CONCLUSIONS

The eGDR index outperforms six other IR indices in predicting CVD, heart disease, and stroke in individuals with CKM syndrome stages 0-3. Its incorporation into predictive models enhances risk stratification and may aid in the early identification of high-risk individuals in this population. Further studies are needed to validate these findings in external cohorts.

Blood-based machine learning classifiers for early diagnosis of gastric cancer via multiple miRNAs

BACKGROUND

Early screening methods for gastric cancer (GC) are lacking; therefore, the disease often progresses to an advanced stage when patients first start to exhibit typical symptoms. Endoscopy and pathological biopsy remain the primary diagnostic approaches, but they are invasive and not yet widely applicable for early population screening. miRNA is a highly conserved type of RNA that exists stably in plasma. Dysfunction of miRNA is linked to tumorigenesis and progression, indicating that individual miRNAs or combinations of multiple miRNAs may serve as potential biomarkers.

AIM

To identify effective plasma miRNA biomarkers and investigate the clinical value of combining multiple miRNAs for early detection of GC.

METHODS

Plasma samples from multiple centres were collected. Differentially expressed genes among healthy controls, early-stage GC patients, and advanced-stage GC patients were identified through small RNA sequencing (sRNA-seq) and validated via real-time quantitative reverse transcription polymerase chain reaction (RT-qPCR). A Wilcoxon signed-rank test was used to investigate the differences in miRNAs. Sequencing datasets of GC serum samples were retrieved from the Gene Expression Omnibus (GEO), ArrayExpress, and The Cancer Genome Atlas databases, and a multilayer perceptron-artificial neural network (MLP-ANN) model was constructed for the key risk miRNAs. The pROC package was used to assess the discriminatory efficacy of the model.

RESULTS

Plasma samples of 107 normal, 71 early GC and 97 advanced GC patients were obtained from three centres, and serum samples of 8443 normal and 1583 GC patients were obtained from the GEO database. The sRNA-seq and RT-qPCR experiments revealed that miR-452-5p, miR-5010-5p, miR-27b-5p, miR-5189-5p, miR-552-5p and miR-199b-5p were significantly increased in early GC patients compared with healthy controls and in advanced GC patients compared with early GC patients (P < 0.05). An MLP-ANN model was constructed for the six key miRNAs. The area under the curve (AUC) within the training cohort was 0.983 [95% confidence interval (CI): 0.980-0.986]. In the two validation cohorts, the AUCs were 0.995 (95%CI: 0.987 to nearly 1.000) and 0.979 (95%CI: 0.972-0.986), respectively.

CONCLUSION

Potential miRNA biomarkers, including miR-452-5p, miR-5010-5p, miR-27b-5p, miR-5189-5p, miR-552-5p and miR-199b-5p, were identified. A GC classifier based on these miRNAs was developed, benefiting early detection and population screening.

Clinician-artificial intelligence collaboration: A win-win solution for efficiency and reliability in atrial fibrillation diagnosis

BACKGROUND

Given the biases and ethical concerns of AI models, the fully automatic diagnosis of diseases in clinical settings is challenging. In contrast, clinician-AI collaboration is considered essential to ensure the validity and reliability of utilizing AI models in clinical practice. However, effective strategies for clinician-AI collaboration remain largely unexplored.

METHODS

This study proposed a three-step general clinician-AI collaboration pipeline aimed at improving disease diagnosis efficiency: first, utilizing large real-world clinical datasets to evaluate and clarify clinicians' diagnostic strengths/weaknesses; second, developing an AI model to complement clinicians' weakness in disease diagnosis; and finally, proposing a clinician-AI collaboration strategy to leverage the strengths of both AI and clinicians. The effectiveness of this pipeline was validated through a study focusing on clinical paroxysmal atrial fibrillation (PAF) detection, utilizing 24-h Holter recordings from over 30,000 patients.

FINDINGS

In PAF detection, clinicians alone required a significant amount of time to identify the data and still overlooked 13.7% of PAF patients but successfully identified all non-atrial fibrillation (AF) patients. Conversely, AI alone rarely missed PAF patients but misidentified 23.3% of non-AF patients as having PAF. After implementing the proposed clinician-AI collaboration strategy, all patients were correctly identified, and clinicians' workload was reduced by 76.7%.

CONCLUSIONS

This study improves both the efficiency and reliability of PAF detection, bridging the gap between AI model development and its clinical application, thereby effectively promoting the application of AI models in clinical AF screening.

FUNDING

This study was supported in part by the National Natural Science Foundation of China.

Recurrence patterns and prediction of survival after recurrence for gallbladder cancer

BACKGROUND

Gallbladder cancer (GBC) is associated with a poor prognosis. Recurrence patterns and their effect on survival remain ill-defined. This study aimed to analyze recurrence patterns and develop a machine learning (ML) model to predict survival after recurrence (SAR) of GBC.

METHODS

Patients who underwent curative-intent resection of GBC between 1999 and 2022 were identified using an international database. An Extreme Gradient Boosting ML model to predict SAR was developed and validated.

RESULTS

Among 348 patients, 110 (31.6%) developed disease recurrence during follow-up. The most common recurrence site was local (29.1%), followed by multiple site (26.4%), liver (21.8%), peritoneal (18.2%), and lung (0.05%). The median SAR was the longest in patients with lung recurrence (36.0 months), followed by those with local recurrence (15.7 months). In contrast, patients with peritoneal (8.9 months), liver (8.5 months), or multiple-site (6.4 months) recurrence had a considerably shorter SAR. Patients with multiple-site recurrence had a worse SAR than individuals with single-site recurrence (6.4 vs 11.10 months, respectively; P =.014). The model demonstrated good performance in the evaluation and bootstrapping cohorts (area under the receiver operating characteristic curve: 71.4 and 71.0, respectively). The most influential variables were American Society of Anesthesiologists classification, local recurrence, receipt of adjuvant chemotherapy, American Joint Committee on Cancer T and N categories, and developing early disease recurrence (<12 months). To enable clinical applicability, an easy-to-use calculator was made available (https://catalano-giovanni.shinyapps.io/SARGB).

CONCLUSION

Except for lung recurrence, SAR for GBC was poor. A subset of patients with less aggressive disease biology may have favorable SAR. ML-based SAR prediction may help individuate candidates for curative re-resection when feasible.

Resveratrol in animal models of pancreatitis and pancreatic cancer: A systematic review with machine learning

BACKGROUND

Resveratrol (RES), a common type of plant polyphenols, has demonstrated promising therapeutic efficacy and safety in animal models of pancreatitis and pancreatic cancer. However, a comprehensive analysis of these data is currently unavailable. This study aimed to systematically review the preclinical evidence regarding RES's effects on animal models of pancreatitis and pancreatic cancer via meta-analyses and optimised machine learning techniques.

METHODS

Animal studies published from inception until June 30th 2024, were systematically retrieved and manually filtrated across databases including PubMed, EMBASE, Web of Science, Ovid MEDLINE, Scopus, and Cochrane Library. Methodological quality of the included studies was evaluated following the SYRCLE's RoB tool. Predefined outcomes included histopathology and relevant biochemical parameters for acute pancreatitis, and tumour weight/tumour volume for pancreatic cancer, comparing treatment and model groups. Pooled effect sizes of the outcomes were calculated using STATA 17.0 software. Machine learning techniques were employed to predict the optimal usage and dosage of RES in pancreatitis models.

RESULTS

A total of 50 studies comprising 33 for acute pancreatitis, 1 chronic pancreatitis, and 16 for pancreatic cancer were included for data synthesis after screening 996 records. RES demonstrated significant improvements on pancreatic histopathology score, pancreatic function parameters (serum amylase and lipase), inflammatory markers (TNF-α, IL-1β, IL-6, and pancreatic myeloperoxidase), oxidative biomarkers (malondialdehyde and superoxide dismutase), and lung injury (lung histopathology and myeloperoxidase) in acute pancreatitis models. In pancreatic cancer models, RES notably reduced tumour weight and volume. Machine learning highlighted tree-structured Parzen estimator-optimised gradient boosted decision tree model as achieving the best performance, identifying course after disease induction, total dosage, single dosage, and total number of doses as critical factors for improving pancreatic histology. Optimal single dosage was 20-105 mg/kg with 3 to 9 doses.

CONCLUSION

This study comprehensively demonstrates the therapeutic effects of RES in mitigating pancreatitis and pancreatic cancer in animal models. Anti-inflammatory, anti-oxidative, and anti-tumour growth properties are potential mechanisms of action for RES.

Methodology and development of a machine learning probability calculator: Data heterogeneity limits ability to predict recurrence after arthroscopic Bankart repair

PURPOSE

The aim of this study was to develop and train a machine learning (ML) algorithm to create a clinical decision support tool (i.e., ML-driven probability calculator) to be used in clinical practice to estimate recurrence rates following an arthroscopic Bankart repair (ABR).

METHODS

Data from 14 previously published studies were collected. Inclusion criteria were (1) patients treated with ABR without remplissage for traumatic anterior shoulder instability and (2) a minimum of 2 years follow-up. Risk factors associated with recurrence were identified using bivariate logistic regression analysis. Subsequently, four ML algorithms were developed and internally validated. The predictive performance was assessed using discrimination, calibration and the Brier score.

RESULTS

In total, 5591 patients underwent ABR with a recurrence rate of 15.4% (n = 862). Age <35 years, participation in contact and collision sports, bony Bankart lesions and full-thickness rotator cuff tears increased the risk of recurrence (all p < 0.05). A single shoulder dislocation (compared to multiple dislocations) lowered the risk of recurrence (p < 0.05). Due to the unavailability of certain variables in some patients, a portion of the patient data had to be excluded before pooling the data set to create the algorithm. A total of 797 patients were included providing information on risk factors associated with recurrence. The discrimination (area under the receiver operating curve) ranged between 0.54 and 0.57 for prediction of recurrence.

CONCLUSION

ML was not able to predict the recurrence following ABR with the current available predictors. Despite a global coordinated effort, the heterogeneity of clinical data limited the predictive capabilities of the algorithm, emphasizing the need for standardized data collection methods in future studies.

LEVEL OF EVIDENCE

Level IV, retrospective cohort study.

Artificial intelligence: A key fulcrum for addressing complex environmental health issues

Environmental health (EH) is a complex and interdisciplinary field dedicated to the examination of environmental behaviours, toxicological effects, health risks, and strategies for mitigating harmful environmental factors. Traditional EH research investigates correlations between risk factors and health outcomes through control variables, but this route is difficult to address complex EH issue. Artificial intelligence (AI) technology not only has accelerated the innovation of the scientific research paradigm but also has become an important tool for solving complex EH problems. However, the in-depth and comprehensive implementation of AI in the field of EH still faces many barriers, such as model generalizability, data privacy protection, algorithm transparency, and regulatory and ethical issues. This review focuses on the compound exposures of EH and explores the potential, challenges, and development directions of AI in four key phases of EH research: (1) data collection, fusion, and management, (2) hazard identification and screening, (3) risk modeling and assessment and (4) EH management. It is not difficult to see that in the future, artificial intelligence technology will inevitably carry out multidimensional simulation of complex exposure factors through multi-mode data fusion, so as to achieve accurate identification of environmental health risks, and eventually become an efficient tool for global environmental health management. This review will help researchers re-examine this strategy and provide a reference for AI to solve complex exposure problems.

Future of Artificial Intelligence-Machine Learning Trends in Pathology and Medicine

Artificial intelligence (AI) and machine learning (ML) are transforming the field of medicine. Health care organizations are now starting to establish management strategies for integrating such platforms (AI-ML toolsets) that leverage the computational power of advanced algorithms to analyze data and to provide better insights that ultimately translate to enhanced clinical decision-making and improved patient outcomes. Emerging AI-ML platforms and trends in pathology and medicine are reshaping the field by offering innovative solutions to enhance diagnostic accuracy, operational workflows, clinical decision support, and clinical outcomes. These tools are also increasingly valuable in pathology research in which they contribute to automated image analysis, biomarker discovery, drug development, clinical trials, and productive analytics. Other related trends include the adoption of ML operations for managing models in clinical settings, the application of multimodal and multiagent AI to utilize diverse data sources, expedited translational research, and virtualized education for training and simulation. As the final chapter of our AI educational series, this review article delves into the current adoption, future directions, and transformative potential of AI-ML platforms in pathology and medicine, discussing their applications, benefits, challenges, and future perspectives.

From challenges and pitfalls to recommendations and opportunities: Implementing federated learning in healthcare

Federated learning holds great potential for enabling large-scale healthcare research and collaboration across multiple centers while ensuring data privacy and security are not compromised. Although numerous recent studies suggest or utilize federated learning based methods in healthcare, it remains unclear which ones have potential clinical utility. This review paper considers and analyzes the most recent studies up to May 2024 that describe federated learning based methods in healthcare. After a thorough review, we find that the vast majority are not appropriate for clinical use due to their methodological flaws and/or underlying biases which include but are not limited to privacy concerns, generalization issues, and communication costs. As a result, the effectiveness of federated learning in healthcare is significantly compromised. To overcome these challenges, we provide recommendations and promising opportunities that might be implemented to resolve these problems and improve the quality of model development in federated learning with healthcare.

Development of a Machine Learning Model to Predict Therapeutic Responses in Laryngopharyngeal Reflux Disease

OBJECTIVES

Laryngopharyngeal reflux disease (LPRD) is a challenging condition requiring effective treatment. Thus, understanding the factors that influence therapeutic response in LPRD is crucial. This study leverages machine learning models to predict the therapeutic responses and identify the key influencing factors in LPRD.

METHODS

Patients with typical LPRD symptoms showing more than one pharyngeal reflux episode on 24-hour multichannel intraluminal impedance (MII)-pH monitoring were collected retrospectively from two independent otolaryngologic clinics. Patients who were prescribed proton pump inhibitor therapy and followed up for at least 2 months were included. Patients who showed a ≥50% decrease in the follow-up reflux symptom index score during treatment periods compared with pre treatment were defined as responders. Among various demographic and 24-hour MII-pH monitoring parameters, features showing the absolute correlation coefficients ≥0.1 with response were selected. Four machine learning models-logistic regression, random forest, support vector machine, and gradient boosting-were applied to the training cohort and assessed in the internal and external validation cohorts.

RESULTS

Patients from two otolaryngologic clinics were assigned to the internal dataset (n = 157) and external dataset (n = 53). All four models showed comparable predictive performances, illustrating their potential utility in clinical decision-making. Among them, the logistic regression model demonstrated the best performance with accuracy and F1 scores of 82.98% and 88.24% in the internal validation cohort and 84.91% and 86.21% in the external validation cohort predicting therapeutic responses in LPRD. Feature importance analysis revealed vital factors, such as proximal total reflux time and weak acid time, influencing therapeutic response, and offering insights into LPRD management.

CONCLUSIONS

This study provides valuable insights into the factors influencing the therapeutic response in LPRD, underscoring the utility of machine learning in refining treatment strategies. Our findings suggest that integrating machine learning models into clinical practice can significantly improve LPRD management.

AI-Derived Blood Biomarkers for Ovarian Cancer Diagnosis: Systematic Review and Meta-Analysis

BACKGROUND

Emerging evidence underscores the potential application of artificial intelligence (AI) in discovering noninvasive blood biomarkers. However, the diagnostic value of AI-derived blood biomarkers for ovarian cancer (OC) remains inconsistent.

OBJECTIVE

We aimed to evaluate the research quality and the validity of AI-based blood biomarkers in OC diagnosis.

METHODS

A systematic search was performed in the MEDLINE, Embase, IEEE Xplore, PubMed, Web of Science, and the Cochrane Library databases. Studies examining the diagnostic accuracy of AI in discovering OC blood biomarkers were identified. The risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies-AI tool. Pooled sensitivity, specificity, and area under the curve (AUC) were estimated using a bivariate model for the diagnostic meta-analysis.

RESULTS

A total of 40 studies were ultimately included. Most (n=31, 78%) included studies were evaluated as low risk of bias. Overall, the pooled sensitivity, specificity, and AUC were 85% (95% CI 83%-87%), 91% (95% CI 90%-92%), and 0.95 (95% CI 0.92-0.96), respectively. For contingency tables with the highest accuracy, the pooled sensitivity, specificity, and AUC were 95% (95% CI 90%-97%), 97% (95% CI 95%-98%), and 0.99 (95% CI 0.98-1.00), respectively. Stratification by AI algorithms revealed higher sensitivity and specificity in studies using machine learning (sensitivity=85% and specificity=92%) compared to those using deep learning (sensitivity=77% and specificity=85%). In addition, studies using serum reported substantially higher sensitivity (94%) and specificity (96%) than those using plasma (sensitivity=83% and specificity=91%). Stratification by external validation demonstrated significantly higher specificity in studies with external validation (specificity=94%) compared to those without external validation (specificity=89%), while the reverse was observed for sensitivity (74% vs 90%). No publication bias was detected in this meta-analysis.

CONCLUSIONS

AI algorithms demonstrate satisfactory performance in the diagnosis of OC using blood biomarkers and are anticipated to become an effective diagnostic modality in the future, potentially avoiding unnecessary surgeries. Future research is warranted to incorporate external validation into AI diagnostic models, as well as to prioritize the adoption of deep learning methodologies.

TRIAL REGISTRATION

PROSPERO CRD42023481232; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023481232.

Applying machine learning algorithms to explore the impact of combined noise and dust on hearing loss in occupationally exposed populations

This study aimed to explore the combined impacts of occupational noise and dust on hearing and extra-auditory functions and identify associated risk factors via machine learning techniques. Data from 14,145 workers (627 with occupational noise-induced hearing loss (ONIHL)) at Hebei Medical Examination Center (2017-2023) were analyzed. Workers with combined exposure and without specific contraindications or other hearing impairment causes were included. Demographic and clinical data were gathered. Chi-square and Mann-Whitney U tests examined variables, and multivariate logistic regression determined ONIHL risk factors. Machine learning algorithms like Logistic Regression and Random Forest were developed, optimized, and evaluated. Results showed significant differences in gender, exposure, blood pressure, smoking, etc. between ONIHL and non-ONIHL groups. Male gender, combined exposure, diastolic blood pressure elevation, smoking, fasting blood glucose elevation, and age were positive predictors, while systolic blood pressure elevation was negative. The logistic model had the highest predictive ability (ROC = 0.714). Subgroup analysis revealed a significant positive correlation in specific subgroups. In summary, combined exposure increased ONIHL risk and affected health. Machine learning effectively predicted ONIHL, but the study had limitations and needed further research.

Machine learning-based progress prediction in accelerated cross-linking for Keratoconus

BACKGROUND

To analyze corneal topographic and biomechanical parameters in keratoconus patients before undergoing accelerated corneal collagen cross-linking (A-CXL) surgery and use machine learning models to identify prognostic factors for disease progression after treatment.

METHODS

This was a retrospective, single-center study on 95 eyes from 69 keratoconus patients (mean age 21.46 ± 7.07 years) undergoing A-CXL, with 3-22 months follow-up. Corneal tomography (Pentacam) and biomechanical measurements (Corvis ST) were performed at baseline and follow-up visits. Changes in the E-stage were used to define progression. LASSO, XGBoost, and random forest machine learning models were applied to identify prognostic factors. A nomogram was developed to predict progression probabilities.

RESULTS

42.1% of eyes showed progression based on E-stage change. Maximal keratometry (Kmax) and index of surface variance (ISV) were significantly higher in the progression group. The nomogram incorporating Kmax and ISV predicted progression better than individual parameters. The progression rate was 51.4% in high-risk eyes versus 16% in low-risk eyes stratified by the nomogram.

CONCLUSIONS

Kmax and ISV are important prognostic factors for keratoconus progression after A-CXL. The nomogram can improve prediction accuracy compared to single parameters. It enables personalized risk assessment to guide treatment decisions.

Deep learning assisted prediction of osteogenic capability of orthopedic implant surfaces based on early cell morphology

The surface modification of titanium (Ti) and its alloys is crucial for improving their osteogenic capability, as their bio-inert nature limits effective osseointegration despite their prevalent use in orthopedic implants. However, these modification methods produce varied surface properties, making it challenging to standardize criteria for assessing the osteogenic capacity of implant surfaces. Additionally, traditional evaluation experiments are time-consuming and inefficient. To overcome these limitations, this study introduced a high-throughput, efficient screening method for assessing the osteogenic capability of implant surfaces based on early cell morphology and deep learning. The Orthopedic Implants-Osteogenic Differentiation Network (OIODNet) was developed using early cell morphology images and corresponding alkaline phosphatase (ALP) activity values from cells cultured on Ti and its alloy surfaces, achieving performance metrics exceeding 0.98 across all six evaluation parameters. Validation through metal-polyphenol network (MPN) coatings and cell experiments demonstrated a strong correlation between OIODNet's predictions and actual ALP activity outcomes, confirming its accuracy in predicting osteogenic potential based on early cell morphology. The Osteogenic Predictor application offers an intuitive tool for predicting the osteogenic capacity of implant surfaces. Overall, this research highlights the potential to accelerate progress at the intersection of artificial intelligence and biomaterials, paving the way for more efficient screening of osteogenic capabilities in orthopedic implants. STATEMENT OF SIGNIFICANCE: By leveraging deep learning, this study introduces the Orthopedic Implants-Osteogenic Differentiation Network (OIODNet), which utilizes early cell morphology data and alkaline phosphatase (ALP) activity values to provide a high-throughput, accurate method for predicting osteogenic capability. With performance metrics exceeding 0.98, OIODNet's accuracy was further validated through experiments involving metal-polyphenol network (MPN) coatings, showing a strong correlation between the model's predictions and experimental outcomes. This research offers a powerful tool for more efficient screening of implant surfaces, marking a transformative step in the integration of artificial intelligence and biomaterials, while opening new avenues for advancing orthopedic implant technologies.

Parallel privacy preservation through partitioning (P4): a scalable data anonymization algorithm for health data

BACKGROUND

Sharing health data holds great potential for advancing medical research but also poses many challenges, including the need to protect people's privacy. One approach to address this is data anonymization, which refers to the process of altering or transforming a dataset to preserve the privacy of the individuals contributing data. To this, privacy models have been designed to measure risks and optimization algorithms can be used to transform data to achieve a good balance between risks reduction and the preservation of the dataset's utility. However, this process is computationally complex and challenging to apply to large datasets. Previously suggested parallel algorithms have been tailored to specific risk models, utility models and transformation methods.

METHODS

We present a novel parallel algorithm that supports a wide range of methods for measuring risks, optimizing utility and transforming data. The algorithm trades data utility for parallelization, by anonymizing partitions of the dataset in parallel. To ensure the correctness of the anonymization process, the algorithm carefully controls the process and if needed rearranges partitions and performs additional transformations.

RESULTS

We demonstrate the effectiveness of our method through an open-source implementation. Our experiments show that our approach can reduce execution times by up to one order of magnitude with minor impacts on output data utility in a wide range of scenarios.

CONCLUSIONS

Our novel P4 algorithm for parallel and distributed data anonymization is, to the best of our knowledge, the first to systematically support a wide variety of privacy, transformation and utility models.

Development and validation of an interpretable machine learning model for predicting the risk of hepatocellular carcinoma in patients with chronic hepatitis B: a case-control study

BACKGROUND

The aim of this study was to develop and internally validate an interpretable machine learning (ML) model for predicting the risk of hepatocellular carcinoma (HCC) in patients with chronic hepatitis B (CHB) infection.

METHODS

We retrospectively collected clinical data from patients with HCC and CHB treated at the Fourth Affiliated Hospital of Guangxi Medical University from January 2022 to December 2022, including demographics, comorbidities, and laboratory parameters. The datasets were randomly divided into a training set (361 cases) and a validation set (155 cases) in a 7:3 ratio. Variables were screened using Least Absolute Shrinkage and Selection Operator (LASSO) and multifactor logistic regression. The prediction model of HCC risk in CHB patients was constructed based on five machine learning models, including logistic regression (LR), K-nearest neighbour (KNN), support vector machine (SVM), random forest (RF) and artificial neural network (ANN). Receiver operating characteristic (ROC) curve, calibration curve and decision curve analysis (DCA) were used to evaluate the predictive performance of the model in terms of identification, calibration and clinical application. The SHapley Additive exPlanation (SHAP) method was used to rank the importance of the features and explain the final model.

RESULTS

Among the five ML models constructed, the RF model has the best performance, and the RF model predicts the risk of HCC in patients with CHB in the training set [AUC: 0.996, 95% confidence interval (CI) (0.991-0.999)] and internal validation set [AUC: 0.993, 95% CI (0.986-1.000)]. It has high AUC, specificity, sensitivity, F1 score and low Brier score. Calibration showed good agreement between observed and predicted risks. The model yielded higher positive net benefits in DCA than when all participants were considered to be at high or low risk, indicating good clinical utility. In addition, the SHAP plot of the RF showed that age, basophil/lymphocyte ratio (BLR), D-Dimer, aspartate aminotransferase/alanine aminotransferase (AST/ALT), γ-glutamyltransferase (GGT) and alpha-fetoprotein (AFP) can help identify patients with CHB who are at high or low risk of developing HCC.

CONCLUSION

ML models can be used as a tool to predict the risk of HCC in patients with CHB. The RF model has the best predictive performance and helps clinicians to identify high-risk patients and intervene early to reduce or delay the occurrence of HCC. However, the model needs to be further improved through large sample studies.

Application of Machine Learning for Patients With Cardiac Arrest: Systematic Review and Meta-Analysis

BACKGROUND

Currently, there is a lack of effective early assessment tools for predicting the onset and development of cardiac arrest (CA). With the increasing attention of clinical researchers on machine learning (ML), some researchers have developed ML models for predicting the occurrence and prognosis of CA, with certain models appearing to outperform traditional scoring tools. However, these models still lack systematic evidence to substantiate their efficacy.

OBJECTIVE

This systematic review and meta-analysis was conducted to evaluate the prediction value of ML in CA for occurrence, good neurological prognosis, mortality, and the return of spontaneous circulation (ROSC), thereby providing evidence-based support for the development and refinement of applicable clinical tools.

METHODS

PubMed, Embase, the Cochrane Library, and Web of Science were systematically searched from their establishment until May 17, 2024. The risk of bias in all prediction models was assessed using the Prediction Model Risk of Bias Assessment Tool.

RESULTS

In total, 93 studies were selected, encompassing 5,729,721 in-hospital and out-of-hospital patients. The meta-analysis revealed that, for predicting CA, the pooled C-index, sensitivity, and specificity derived from the imbalanced validation dataset were 0.90 (95% CI 0.87-0.93), 0.83 (95% CI 0.79-0.87), and 0.93 (95% CI 0.88-0.96), respectively. On the basis of the balanced validation dataset, the pooled C-index, sensitivity, and specificity were 0.88 (95% CI 0.86-0.90), 0.72 (95% CI 0.49-0.95), and 0.79 (95% CI 0.68-0.91), respectively. For predicting the good cerebral performance category score 1 to 2, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.86 (95% CI 0.85-0.87), 0.72 (95% CI 0.61-0.81), and 0.79 (95% CI 0.66-0.88), respectively. For predicting CA mortality, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.85 (95% CI 0.82-0.87), 0.83 (95% CI 0.79-0.87), and 0.79 (95% CI 0.74-0.83), respectively. For predicting ROSC, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.77 (95% CI 0.74-0.80), 0.53 (95% CI 0.31-0.74), and 0.88 (95% CI 0.71-0.96), respectively. In predicting CA, the most significant modeling variables were respiratory rate, blood pressure, age, and temperature. In predicting a good cerebral performance category score 1 to 2, the most significant modeling variables in the in-hospital CA group were rhythm (shockable or nonshockable), age, medication use, and gender; the most significant modeling variables in the out-of-hospital CA group were age, rhythm (shockable or nonshockable), medication use, and ROSC.

CONCLUSIONS

ML represents a currently promising approach for predicting the occurrence and outcomes of CA. Therefore, in future research on CA, we may attempt to systematically update traditional scoring tools based on the superior performance of ML in specific outcomes, achieving artificial intelligence-driven enhancements.

TRIAL REGISTRATION

PROSPERO International Prospective Register of Systematic Reviews CRD42024518949; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=518949.

Prediction of postpartum depression in women: development and validation of multiple machine learning models

BACKGROUND

Postpartum depression (PPD) is a significant public health issue. This study aimed to develop and validate machine learning (ML) models using biopsychosocial predictors to predict the risk of PPD for perinatal women and to provide several risk assessment tools for the early detection of PPD.

METHODS

Candidate predictors, including history of mental illness and demographic, psychosocial, and physiological factors, were obtained from 1138 perinatal women between August 2021 and August 2022. The primary outcome of PPD was measured with the Edinburgh Postnatal Depression Scale at 6 weeks postpartum. Seven feature selection methods and six ML algorithms were employed to develop models, and their prediction performances were compared.

RESULTS

A total of 11 potential predictive factors associated with PPD were identified and subsequently used to construct prenatal and postpartum predictive models for PPD. The cross-validation results showed that the models built on logistic regression (LR) [area under the curve (AUC): 0.801, 0.858] and artificial neural network (ANN) (AUC: 0.787, 0.844) algorithms exhibited the best prediction performance. In contrast to the prenatal models, the addition of postpartum predictors (primary caregiver and mother-in-law's care) remarkably improved the predictive performance of the postpartum models. The risk-stratification score, the nomogram, and the Shapley additive explanation were used to visualize and interpret the risk prediction model for predicting PPD in the early stage.

CONCLUSIONS

The LR and ANN models achieved the best predictive performances. Applying these models and risk assessment tools to early predict and screen PPD has several implications for public health.

Prediction of malnutrition in kids by integrating ResNet-50-based deep learning technique using facial images

In recent times, severe acute malnutrition (SAM) in India is considered a serious issue as per UNICEF 2022 records. In that record, 35.5% of children under age 5 are stunted, 19.3% are wasted, and 32% are underweight. Malnutrition, defined as these three conditions, affects 5.7 million children globally. This research utilizes an artificial intelligence-based image segmentation technique to predict malnutrition in children. The primary goal of this research is to use a deep learning model to eliminate the need for multiple manual diagnostic tests and simplify the prediction of malnutrition in kids. The traditional model uses text-based data and takes more time with continuous monitoring of kids by analysing body mass index (BMI) over different periods. Children in rural areas often miss medical expert appointments, and a lack of knowledge among parents can lead to severe malnutrition. The aim of the proposed system is to eliminate the need for manual blood tests and regular visits to medical experts. This study uses the ResNet-50 deep learning model's built-in shortcut connection to solve the image-based vanishing gradient problem. This makes training more efficient for image segmentation tasks in predicting malnutrition. The model is 98.49% accurate in predicting the kids who are malnourished among the kids who are healthy. It is evident from the results that the proposed system serves better than other deep learning models, such as XG Boost (75.29% accuracy), VGG 16 (94% accuracy), Xception (95.41% accuracy), and MobileNet (92.42% accuracy). Hence, the proposed technique is effective in detecting malnutrition and diagnose it earlier, without using predictive analysis function or advice from the medical experts.

Deep-Learning-Based Analysis of Electronic Skin Sensing Data

E-skin is an integrated electronic system that can mimic the perceptual ability of human skin. Traditional analysis methods struggle to handle complex e-skin data, which include time series and multiple patterns, especially when dealing with intricate signals and real-time responses. Recently, deep learning techniques, such as the convolutional neural network, recurrent neural network, and transformer methods, provide effective solutions that can automatically extract data features and recognize patterns, significantly improving the analysis of e-skin data. Deep learning is not only capable of handling multimodal data but can also provide real-time response and personalized predictions in dynamic environments. Nevertheless, problems such as insufficient data annotation and high demand for computational resources still limit the application of e-skin. Optimizing deep learning algorithms, improving computational efficiency, and exploring hardware-algorithm co-designing will be the key to future development. This review aims to present the deep learning techniques applied in e-skin and provide inspiration for subsequent researchers. We first summarize the sources and characteristics of e-skin data and review the deep learning models applicable to e-skin data and their applications in data analysis. Additionally, we discuss the use of deep learning in e-skin, particularly in health monitoring and human-machine interactions, and we explore the current challenges and future development directions.

Predicting the complexity of minimally invasive liver resection for hepatocellular carcinoma using machine learning

BACKGROUND

Despite technical advancements, minimally invasive liver surgery (MILS) for hepatocellular carcinoma (HCC) remains challenging. Nonetheless, effective tools to assess MILS complexity are still lacking. Machine learning (ML) models could improve the accuracy of such tools.

METHODS

Patients who underwent curative-intent MILS for HCC were identified using an international database. An XGBoost ML model was developed to predict surgical complexity using clinical and radiological characteristics.

RESULTS

Among 845 patients, 186 (22.0 %) were classified as high-risk patients. In this subgroup, median Charlson Comorbidity Index (CCI) (5.0, IQR 3.0-7.0 vs. 2.0, IQR 2.0-5.0, p < 0.001) and tumor burden score (TBS) (median 4.12, IQR 3.0-5.1 vs. 4.22, IQR 3.2-7.1, p < 0.001) were higher. The model was able to effectively predict complexity of surgery in both the training and testing cohorts with high discriminating power (ROC-AUC: 0.86, 95%CI 0.82-0.89 vs. 0.73, 95%CI 0.65-0.81). The most influential variables were CCI, TBS, BMI, extent of resection, and sex. Patients predicted to have a complex surgery were more likely to develop severe complications (OR 4.77, 95%CI 1.82-13.9, p = 0.002). An easy-to-use calculator was developed.

CONCLUSION

Preoperative ML-prediction of complex MILS for HCC may improve preoperative planning, resource allocation, and patient outcomes.

Machine-learning models to predict serious adverse hospitalization events after ACS

OBJECTIVE

We developed a risk stratification model to predict serious adverse hospitalization events (mortality, cardiac shock, cardiac arrest) (SAHE) after acute coronary syndrome (ACS) based on machine-learning models and logistic regression model.

METHODS

This cohort study is based on the CCC-ACS project. The primary efficacy outcomes were SAHE. Clinical prediction models were established based on five machine-learning (XGBoost, RF, MLP, KNN, and stacking model) and logistic regression models.

RESULTS

Among the 112 363 patients in the study, age (55-65 years: OR: 1.392; 95%CI: 1.212-1.600; 65-75 years: OR: 1.878; 95%CI: 1.647-2.144; ≥75 year: OR: 2.976; 95%CI: 2.615-3.393), history of diabetes mellitus (OR: 1.188; 95%CI: 1.083-1.302), history of renal failure (OR: 1.645; 95%CI: 1.311-2.044), heart rate (60-100 beats/min: OR: 0.468; 95%CI: 0.409-0.536; ≥100 beats/min: OR: 0.540; 95%CI: 0.454-0.643), shock index (0.4-0.8: OR: 1.796; 95%CI: 1.440-2.264; ≥0.8: OR: 5.883; 95%CI: 4.619-7.561), KILLIP (II: OR: 1.171; 95%CI: 1.048-1.306; III: OR: 1.696; 95%CI: 1.469-1.952; IV: OR: 7.811; 95%CI: 7.023-8.684), and cardiac arrest at admission (OR: 12.507; 95%CI: 10.757-14.530) were independent predictors of severe adverse hospitalization events for ACS patients. In several machine-learning models, RF (AUC: 0.817; 95%CI: 0.808-0.826) and XGBoost (AUC: 0.816; 95%CI: 0.807-0.825) also showed good discrimination in the training set, which ranked the first two positions. They also presented good accuracy and the best clinical benefits in the decision curve analysis. In addition, logistic regression was able to discriminate the SAHE (AUC: 0.816; 95%CI: 0.807-0.825) and performed the best prediction accuracy (0.822; 95%CI: 0.822-0.822) compared to several machine-learning models. Model calibration and decision curve analysis showed these prediction models have similar predictive performance. Based on these findings, we developed two CCC-ACS In-hospital Major Adverse Events Risk Scores and its online calculator. One is based on machine-learning model (https://ccc-acs-sae-3-xcnjsvoccusjwkfhfthh44.streamlit.app/), and another is based on logistic regression model (https://ccc-acs-sae-logistic-9te57ylnq3kazkeuyc7dub.streamlit.app/), offering a validated tool to predict survival for patients with ACS during hospitalization.

CONCLUSIONS

Machine-learning-based approaches for identifying predictors of SAHE after an ACS were feasible and practical. Based on this, we developed two online risk prediction websites for clinicians' decision-making. The CCC-ACS-MSAE score showed accurate discriminative capabilities for predicting severe adverse hospitalization events and might help guide clinical decision-making. Key messages: Three research questions and three bullet points What is already known on this topic? Observational studies have identified risk factors for in-hospital death in patients with acute coronary syndromes (ACS). However, the real-world results of a large sample in China still need to be further explored. What does this study add? Machine-learning-based approaches for identifying predictors of SAHE after an ACS were feasible and practical. Based on these findings, we developed two CCC-ACS In-hospital Major Adverse Events Risk Scores and its online calculator. One is based on machine-learning model (https://ccc-acs-sae-3-xcnjsvoccusjwkfhfthh44.streamlit.app/), and another is based on logistic regression model (https://ccc-acs-sae-logistic-9te57ylnq3kazkeuyc7dub.streamlit.app/), offering a validated tool to predict survival for patients with ACS during hospitalization. How this study might affect research, practice, or policy? Early identification of high-risk ACS patients will help reduce in-hospital deaths and improve the prognosis of ACS patients.

Interpretable machine learning model for early morbidity risk prediction in patients with sepsis-induced coagulopathy: a multi-center study

Background

Sepsis-induced coagulopathy (SIC) is a complex condition characterized by systemic inflammation and coagulopathy. This study aimed to develop and validate a machine learning (ML) model to predict SIC risk in patients with sepsis.

Methods

Patients with sepsis admitted to the intensive care unit (ICU) between March 1, 2021, and March 1, 2024, at Hebei General Hospital and Handan Central Hospital (East District) were retrospectively included. Patients were categorized into SIC and non-SIC groups. Data were split into training (70%) and testing (30%) sets. Additionally, for temporal validation, patients with sepsis admitted between March 1, 2024, and October 31, 2024, at Hebei General Hospital were included. Feature selection was performed using least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression. Nine ML algorithms were tested, and model performance was assessed using receiver operating characteristic curve (ROC) analysis, including area under the curve (AUC), calibration curves, and decision curve analysis (DCA). The SHaply Additive Explanations (SHAP) algorithm was used to interpret the best-performing model and visualize key predictors.

Results

Among 847 patients with sepsis, 480 (56.7%) developed SIC. The random forest (RF) model with eight variables performed best, achieving AUCs of 0.782 [95% confidence interval (CI): 0.745, 0.818] in the training set, 0.750 (95% CI: 0.690, 0.809) in the testing set, and 0.784 (95% CI: 0.711, 0.857) in the validation set. Key predictors included activated partial thromboplastin time, lactate, oxygenation index, and total protein.

Conclusions

This ML model reliably predicts SIC risk. SHAP enhances interpretability, supporting early, individualized interventions to improve outcomes in patients with sepsis.

Assessing prospective molecular biomarkers and functional pathways in severe asthma based on a machine learning method and bioinformatics analyses

BACKGROUND

Severe asthma, which differs significantly from typical asthma, involves specific molecular biomarkers that enhance our understanding and diagnostic capabilities. The objective of this study is to assess the biological processes underlying severe asthma and to detect key molecular biomarkers.

METHODS

We used Weighted Gene Co-Expression Network Analysis (WGCNA) to detect hub genes in the GSE143303 dataset and indicated their functions and regulatory mechanisms using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and Gene Ontology (GO) annotations. In the GSE147878 dataset, we used Gene Set Enrichment Analysis (GSEA) to determine the regulatory directions of gene sets. We detected differentially expressed genes in the GSE143303 and GSE64913 datasets, constructed a Least Absolute Shrinkage and Selection Operator (LASSO) regression model, and validated the model using the GSE147878 dataset and real-time quantitative PCR (RT-qPCR) to confirm the molecular biomarkers.

RESULTS

Using WGCNA, we discovered modules that were strongly correlated with clinical features, specifically the purple module (r = 0.53) and the midnight blue module (r = -0.65). The hub genes within these modules were enriched in pathways related to mitochondrial function and oxidative phosphorylation. GSEA in the GSE147878 dataset revealed significant enrichment of upregulated gene sets associated with oxidative phosphorylation and downregulated gene sets related to asthma. We discovered 12 commonly regulated genes in the GSE143303 and GSE64913 datasets and developed a LASSO regression model. The model corresponding to lambda.min selected nine genes, including TFCP2L1, KRT6A, FCER1A, and CCL5, which demonstrated predictive value. These genes were significantly upregulated or under expressed in severe asthma, as validated by RT-qPCR.

CONCLUSION

Mitochondrial abnormalities affecting oxidative phosphorylation play a critical role in severe asthma. Key molecular biomarkers like TFCP2L1, KRT6A, FCER1A, and CCL5, are essential for detecting severe asthma. This research significantly enhances the understanding and diagnosis of severe asthma.

Leveraging artificial intelligence and machine learning to accelerate discovery of disease-modifying therapies in type 1 diabetes

Progress in developing therapies for the maintenance of endogenous insulin secretion in, or the prevention of, type 1 diabetes has been hindered by limited animal models, the length and cost of clinical trials, difficulties in identifying individuals who will progress faster to a clinical diagnosis of type 1 diabetes, and heterogeneous clinical responses in intervention trials. Classic placebo-controlled intervention trials often include monotherapies, broad participant populations and extended follow-up periods focused on clinical endpoints. While this approach remains the 'gold standard' of clinical research, efforts are underway to implement new approaches harnessing the power of artificial intelligence and machine learning to accelerate drug discovery and efficacy testing. Here, we review emerging approaches for repurposing agents used to treat diseases that share pathogenic pathways with type 1 diabetes and selecting synergistic combinations of drugs to maximise therapeutic efficacy. We discuss how emerging multi-omics technologies, including analysis of antigen processing and presentation to adaptive immune cells, may lead to the discovery of novel biomarkers and subsequent translation into antigen-specific immunotherapies. We also discuss the potential for using artificial intelligence to create 'digital twin' models that enable rapid in silico testing of personalised agents as well as dose determination. To conclude, we discuss some limitations of artificial intelligence and machine learning, including issues pertaining to model interpretability and bias, as well as the continued need for validation studies via confirmatory intervention trials.

Omics in acute-on-chronic liver failure

Acute-on-chronic liver failure (ACLF) is a critical syndrome that develops in patients with chronic liver disease and is characterized by acute decompensation, single- or multiple-organ failure and high short-term mortality. Over the past few decades, ACLF has been progressively recognized as an independent clinical entity, and several criteria and prognostic scores have been proposed and validated by different scientific societies. However, controversies still exist in some aspects across regions, which mainly involve whether the definition of underlying liver diseases should include cirrhosis and non-cirrhosis. The pathophysiology of ACLF is complicated and remains unclear, although accumulating evidence based on different aetiologies of ACLF shows that it is closely associated with intense systemic inflammation and immune-metabolism disorder, which result in mitochondrial dysfunction and microenvironment imbalance, leading to disease development and organ failure. In-depth insight into the biological pathways involved in the mechanisms of ACLF and potential mechanistic targets that improve patient survival still needs to be investigated. Omics-based analytical techniques, including genomics, transcriptomics, proteomics, metabolomics and microbiomes, have developed rapidly and can offer novel insights into the essential pathophysiologic process of ACLF. In this paper, we briefly reviewed and summarized the current knowledge and recent advances in the definitions, criteria and prognostic assessments of ACLF; we also described the omics techniques and how omics-based analyses have been applied to investigate and characterize the biological mechanisms of ACLF and identify potential predictive biomarkers and therapeutic targets for ACLF. We also outline the challenges, future directions and limitations presented by omics-based analyses in clinical ACLF research.

Development and validation of an ultrasound-based interpretable machine learning model for the classification of ≤3 cm hepatocellular carcinoma: a multicentre retrospective diagnostic study

Background

Our study aimed to develop a machine learning (ML) model utilizing grayscale ultrasound (US) to distinguish ≤3 cm small hepatocellular carcinoma (sHCC) from non-HCC lesions.

Methods

A total of 1052 patients with 1058 liver lesions ≤3 cm from 55 hospitals were collected between May 2017 and June 2021, and 756 liver lesions were randomly allocated into train and internal validation cohorts at a 8:2 ratio for the development and evaluation of ML models based on multilayer perceptron (MLP) and extreme gradient boosting (XGBoost) methods (ModelU utilizing US imaging features; ModelUR adding US radiomics features; ModelURC employing clinical features further). The diagnostic performance of three models was assessed in external validation cohort (312 liver lesions from 14 hospitals). The diagnostic efficacy of the optimal model was compared to that of radiologists in external validation cohort. The SHapley Additive exPlanations (SHAP) method was employed to interpret the optimal ML model by ranking feature importance. The study was registered at ClinicalTrials.gov (NCT03871140).

Findings

ModelURC based XGBoost showed the best performance (AUC = 0.934; 95% CI: 0.894-0.974) in the internal validation cohort. In the external validation cohort, ModelURC also achieved optimal AUC (AUC = 0.899, 95% CI: 0.861-0.931). Upon conducting a subgroup analysis, no statistically significant differences were observed in the diagnostic performance of the ModelURC neither between tumor sizes of ≤2.0 cm and 2.1-3.0 cm nor across different HCC risk stratifications. ModelURC exhibited superior ability compared to all radiologists and ModelURC assistance significantly improved the diagnostic AUC for all radiologists (all P < 0.0001).

Interpretation

A diagnostic model for sHCC was developed and validated using ML and grayscale US from large cohorts. This model significantly improved the diagnostic performance of grayscale US for sHCC compared with experts.

Funding

This work was supported by National Key Research and Development Program of China (2022YFC2405500), Major Research Program of the National Natural Science Foundation of China (92159305), National Science Fund for Distinguished Young Scholars (82325027), Key project of National Natural Science Foundation of China (82030047), Military Fund for Geriatric Diseases (20BJZ42), National Natural Science Foundation of China Special Program (82441011). National Natural Science Foundation of China (82402280), National Natural Science Foundation of China (32171363), Key Research and Development Program for Social Development of Yunnan Science and Technology Department (202403AC100014).