Comparison of MPL-ANN and PLS-DA models for predicting the severity of patients with acute pancreatitis: An exploratory study

Constructing a prediction model for acute pancreatitis severity based on liquid neural network

Acute pancreatitis (AP) is a common disease, and severe acute pancreatitis (SAP) has a high morbidity and mortality rate. Early recognition of SAP is crucial for prognosis. This study aimed to develop a novel liquid neural network (LNN) model for predicting SAP. This study retrospectively analyzed the data of AP patients admitted to the Second Affiliated Hospital of Guilin Medical University between January 2020 and June 2024. Data imbalance was dealt with by data preprocessing and using the synthetic minority oversampling technique (SMOTE). A new feature selection method was designed to optimize model performance. Logistic regression (LR), decision tree (DCT), random forest (RF), Extreme Gradient Boosting (XGBoost), and LNN models were built. The model's performance was evaluated by calculating the area under the receiver operating characteristic (ROC) curve (AUC) and other statistical metrics. In addition, SHapley Additive exPlanations (SHAP) analysis was used to interpret the prediction results of the LNN model. The LNN model performed best in predicting AP severity, with an AUC value of 0.9659 and accuracy, precision, recall, F1 score, and specificity higher than 0.90. SHAP analysis revealed key predictors, such as calcium level, amylase activity, and percentage of basophils, which were strongly associated with AP severity. As an emerging machine learning tool, the LNN model has demonstrated excellent performance and potential in AP severity prediction. The results of this study support the idea that LNN models can be applied to early severity assessment of AP patients in a clinical setting, which can help optimize treatment plans and improve patient prognosis.

Chemometrics and analytical blank on the at-line monitoring of Zika-VLP production using near-infrared spectroscopy

The Zika disease caused by the Zika virus was declared a Public Health Emergency by the World Health Union (WHO), with microcephaly as the most critical consequence. Aiming to reduce the spread of the virus, biopharmaceutical organizations invest in vaccine research and production, based on multiple platforms. A crescent vaccine production approach is based on virus-like particles (VLP), for not having genetic material in its composition, hypoallergenic and non-mutant character. For bioprocess, it is essential to have means of real-time monitoring, which can be assessed using process analysis techniques such as Near-infrared (NIR) spectroscopy, that can be combined with chemometric methods, like Partial-Least Squares (PLS) and Artificial Neural Networks (ANN) for prediction of biochemical variables. This work proposes a biochemical Zika VLP upstream production at-line monitoring model using NIR spectroscopy comparing sampling conditions (with or without cells), analytical blank (air, ultrapure water), and spectra pre-processing approaches. Seven experiments in a benchtop bioreactor using recombinant baculovirus/Sf9 insect cell platform in serum-free medium were performed to obtain biochemical and spectral data for chemometrics modeling (PLS and ANN), composed by a random data split (80 % calibration, 20 % validation) for cross-validation of the PLS models and 70 % training, 15 % testing, 15 % validation for ANN. The best models generated in the present work presented an average absolute error of 1.59 × 105 cell/mL for density of viable cells, 2.37 % for cell viability, 0.25 g/L for glucose, 0.007 g/L for lactate, 0.138 g/L for glutamine, 0.18 g/L for glutamate, 0,003 g/L for ammonium, and 0.014 g/L for potassium.

Development and validation of an explainable machine learning model for mortality prediction among patients with infected pancreatic necrosis

Background

Infected pancreatic necrosis (IPN) represents a severe complication of acute pancreatitis, commonly linked with mortality rates ranging from 15% to 35%. However, the present mortality prediction tools for IPN are limited and lack sufficient sensitivity and specificity. This study aims to develop and validate an explainable machine learning (ML) model for death prediction among patients with IPN.

Methods

We performed a prospective cohort study of 344 patients with IPN consecutively enrolled from a large Chinese tertiary hospital from January 2011 to January 2023. Ten ML models were developed to predict 90-day mortality in these patients. A benchmarking test, involving nested resampling, automatic hyperparameter tuning and random search techniques, was conducted to select the ML model. Sequential forward selection method was employed to select the optimal feature subset from 31 candidate subsets to simplify the model and maximize predictive performance. The final model was internally validated with the 1000 bootstrap method and externally validated using an independent cohort of 132 patients with IPN retrospectively collected from another Chinese tertiary hospital from January 2018 to January 2023. The SHapley Additive exPlanations (SHAP) method was employed to interpret the model in terms of features importance and features effect. The final model constructed with optimal feature subset was deployed as an interactive web-based Shiny app.

Findings

Random survival forest (RSF) model showed the best predictive performance than other 9 ML models (internal validation, C-index = 0.863 [95% CI: 0.854-0.875]; external validation, C-index = 0.857 [95% CI: 0.850-0.865]). Multiple organ failure, Acute Physiology and Chronic Health Examination II (APACHE II) score ≥20, duration of organ failure ≥21 days, bloodstream infection, time from onset to first intervention <30 days, Bedside Index of Severity in Acute Pancreatitis score ≥3, critical acute pancreatitis, age ≥ 50 years, and hemorrhage were 9 most important features associated with mortality. Furthermore, SHAP algorithm revealed insightful nonlinear interactive associations between important predictors and mortality, identifying 9 features pairs with high interaction SHAP value and clinical significance. Two interactive web-based Shiny apps were developed to enhance clinical practicability: https://rsfmodels.shinyapps.io/IPN_app/ for cases where the APACHE II score was available and https://rsfmodels.shinyapps.io/IPNeasy/ for cases where it was not.

Interpretation

An explainable ML model for death prediction among IPN patients was feasible and effective, suggesting its superior potential in guiding clinical management and improving patient outcomes. Two publicly accessible web tools generated for the optimized model facilitated its utility in clinical settings.

Funding

The Natural Science Foundation of Hunan Province (2023JJ30885), Postdoctoral Fellowship Program of CPSF (GZB20230872), The Youth Science Foundation of Xiangya Hospital (2023Q13), The Project Program of National Clinical Research Center for Geriatric Disorders of Xiangya Hospital (2021LNJJ19).

A systematic review of machine learning-based prognostic models for acute pancreatitis: Towards improving methods and reporting quality

BACKGROUND

An accurate prognostic tool is essential to aid clinical decision-making (e.g., patient triage) and to advance personalized medicine. However, such a prognostic tool is lacking for acute pancreatitis (AP). Increasingly machine learning (ML) techniques are being used to develop high-performing prognostic models in AP. However, methodologic and reporting quality has received little attention. High-quality reporting and study methodology are critical for model validity, reproducibility, and clinical implementation. In collaboration with content experts in ML methodology, we performed a systematic review critically appraising the quality of methodology and reporting of recently published ML AP prognostic models.

METHODS/FINDINGS

Using a validated search strategy, we identified ML AP studies from the databases MEDLINE and EMBASE published between January 2021 and December 2023. We also searched pre-print servers medRxiv, bioRxiv, and arXiv for pre-prints registered between January 2021 and December 2023. Eligibility criteria included all retrospective or prospective studies that developed or validated new or existing ML models in patients with AP that predicted an outcome following an episode of AP. Meta-analysis was considered if there was homogeneity in the study design and in the type of outcome predicted. For risk of bias (ROB) assessment, we used the Prediction Model Risk of Bias Assessment Tool. Quality of reporting was assessed using the Transparent Reporting of a Multivariable Prediction Model of Individual Prognosis or Diagnosis-Artificial Intelligence (TRIPOD+AI) statement that defines standards for 27 items that should be reported in publications using ML prognostic models. The search strategy identified 6,480 publications of which 30 met the eligibility criteria. Studies originated from China (22), the United States (4), and other (4). All 30 studies developed a new ML model and none sought to validate an existing ML model, producing a total of 39 new ML models. AP severity (23/39) or mortality (6/39) were the most common outcomes predicted. The mean area under the curve for all models and endpoints was 0.91 (SD 0.08). The ROB was high for at least one domain in all 39 models, particularly for the analysis domain (37/39 models). Steps were not taken to minimize over-optimistic model performance in 27/39 models. Due to heterogeneity in the study design and in how the outcomes were defined and determined, meta-analysis was not performed. Studies reported on only 15/27 items from TRIPOD+AI standards, with only 7/30 justifying sample size and 13/30 assessing data quality. Other reporting deficiencies included omissions regarding human-AI interaction (28/30), handling low-quality or incomplete data in practice (27/30), sharing analytical codes (25/30), study protocols (25/30), and reporting source data (19/30).

CONCLUSIONS

There are significant deficiencies in the methodology and reporting of recently published ML based prognostic models in AP patients. These undermine the validity, reproducibility, and implementation of these prognostic models despite their promise of superior predictive accuracy.

REGISTRATION

Research Registry (reviewregistry1727).

Improved outcome prediction in acute pancreatitis with generated data and advanced machine learning algorithms

OBJECTIVES

Traditional scoring systems have been widely used to predict acute pancreatitis (AP) severity but have limitations in predictive accuracy. This study investigates the use of machine learning (ML) algorithms to improve predictive accuracy in AP.

METHODS

A retrospective study was conducted using data from 101 AP patients in a tertiary hospital in Türkiye. Data were preprocessed, and synthetic data were generated with Gaussian noise addition and balanced with the ADASYN algorithm, resulting in 250 cases. Supervised ML models, including random forest (RF) and XGBoost (XGB), were trained, tested, and validated against traditional clinical scores (Ranson's, modified Glasgow, and BISAP) using area under the curve (AUC), F1 score, and recall.

RESULTS

RF outperformed XGB with an AUC of 0.89, F1 score of 0.82, and recall of 0.82. BISAP showed balanced performance (AUC = 0.70, F1 = 0.44, and recall = 0.85), whereas the Glasgow criteria had the highest recall but lower precision (AUC = 0.70, F1 = 0.38, and recall = 0.95). Ranson's admission criteria were the least effective (AUC = 0.53, F1 = 0.42, and recall = 0.39), probable because it lacked the 48th h features.

CONCLUSION

ML models, especially RF, significantly outperform traditional clinical scores in predicting adverse outcomes in AP, suggesting that integrating ML into clinical practice could improve prognostic assessments.

Raman laser intensity and sample clarification on biochemical monitoring over Zika-VLP upstream stages

In the current biopharmaceutical scenario, constant bioprocess monitoring is crucial for the quality and integrity of final products. Thus, process analytical techniques, such as those based on Raman spectroscopy, have been used as multiparameter tracking methods in pharma bioprocesses, which can be combined with chemometric tools, like Partial Least Squares (PLS) and Artificial Neural Networks (ANN). In some cases, applying spectra pre-processing techniques before modeling can improve the accuracy of chemometric model fittings to observed values. One of the biological applications of these techniques could have as a target the virus-like particles (VLP), a vaccine production platform for viral diseases. A disease that has drawn attention in recent years is Zika, with large-scale production sometimes challenging without an appropriate monitoring approach. This work aimed to define global models for Zika VLP upstream production monitoring with Raman considering different laser intensities (200 mW and 495 mW), sample clarification (with or without cells), spectra pre-processing approaches, and PLS and ANN modeling techniques. Six experiments were performed in a benchtop bioreactor to collect the Raman spectral and biochemical datasets for modeling calibration. The best models generated presented a mean absolute error and mean relative error respectively of 3.46 × 105 cell/mL and 35 % for viable cell density (Xv); 4.1 % and 5 % for cell viability (CV); 0.245 g/L and 3 % for glucose (Glc); 0.006 g/L and 18 % for lactate (Lac); 0.115 g/L and 26 % for glutamine (Gln); 0.132 g/L and 18 % for glutamate (Glu); 0.0029 g/L and 3 % for ammonium (NH4+); and 0.0103 g/L and 2 % for potassium (K+). Sample without conditioning (with cells) improved the models' adequacy, except for Glutamine. ANN better predicted CV, Gln, Glu, and K+, while Xv, Glc, Lac, and NH4+ presented no statistical difference between the chemometric tools. For most of the assessed experimental parameters, there was no statistical need for spectra pre-filtering, for which the models based on the raw spectra were selected as the best ones. Laser intensity impacts quality model predictions in some parameters, Xv, Gln, and K+ had a better performance with 200 mW of intensity (for PLS, ANN, and ANN, respectively), for CV the 495 mW laser intensity was better (for PLS), and for the other biochemical variables, the use of 200 or 495 mW did not impact model fitting adequacy.

Predictive value of machine learning for the severity of acute pancreatitis: A systematic review and meta-analysis

Background

Predicting the severity of acute pancreatitis (AP) early poses a challenge in clinical practice. While there are well-established clinical scoring tools, their actual predictive performance remains uncertain. Various studies have explored the application of machine-learning methods for early AP prediction. However, a more comprehensive evidence-based assessment is needed to determine their predictive accuracy. Hence, this systematic review and meta-analysis aimed to evaluate the predictive accuracy of machine learning in assessing the severity of AP.

Methods

PubMed, EMBASE, Cochrane Library, and Web of Science were systematically searched until December 5, 2023. The risk of bias in eligible studies was assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST). Subgroup analyses, based on different machine learning types, were performed. Additionally, the predictive accuracy of mainstream scoring tools was summarized.

Results

This systematic review ultimately included 33 original studies. The pooled c-index in both the training and validation sets was 0.87 (95 % CI: 0.84-0.89) and 0.88 (95 % CI: 0.86-0.90), respectively. The sensitivity in the training set was 0.81 (95 % CI: 0.77-0.84), and in the validation set, it was 0.79 (95 % CI: 0.71-0.85). The specificity in the training set was 0.84 (95 % CI: 0.78-0.89), and in the validation set, it was 0.90 (95 % CI: 0.86-0.93). The primary model incorporated was logistic regression; however, its predictive accuracy was found to be inferior to that of neural networks, random forests, and xgboost. The pooled c-index of the APACHE II, BISAP, and Ranson were 0.74 (95 % CI: 0.68-0.80), 0.77 (95 % CI: 0.70-0.85), and 0.74 (95 % CI: 0.68-0.79), respectively.

Conclusions

Machine learning demonstrates excellent accuracy in predicting the severity of AP, providing a reference for updating or developing a straightforward clinical prediction tool.

Artificial intelligence: Applications in critical care gastroenterology

Gastrointestinal (GI) complications frequently necessitate intensive care unit (ICU) admission. Additionally, critically ill patients also develop GI complications requiring further diagnostic and therapeutic interventions. However, these patients form a vulnerable group, who are at risk for developing side effects and complications. Every effort must be made to reduce invasiveness and ensure safety of interventions in ICU patients. Artificial intelligence (AI) is a rapidly evolving technology with several potential applications in healthcare settings. ICUs produce a large amount of data, which may be employed for creation of AI algorithms, and provide a lucrative opportunity for application of AI. However, the current role of AI in these patients remains limited due to lack of large-scale trials comparing the efficacy of AI with the accepted standards of care.

Machine learning-based decision tool for selecting patients with idiopathic acute pancreatitis for endosonography to exclude a biliary aetiology

BACKGROUND

Biliary microlithiasis/sludge is detected in approximately 30% of patients with idiopathic acute pancreatitis (IAP). As recurrent biliary pancreatitis can be prevented, the underlying aetiology of IAP should be established.

AIM

To develop a machine learning (ML) based decision tool for the use of endosonography (EUS) in pancreatitis patients to detect sludge and microlithiasis.

METHODS

We retrospectively used routinely recorded clinical and laboratory parameters of 218 consecutive patients with confirmed AP admitted to our tertiary care hospital between 2015 and 2020. Patients who did not receive EUS as part of the diagnostic work-up and whose pancreatitis episode could be adequately explained by other causes than biliary sludge and microlithiasis were excluded. We trained supervised ML classifiers using H2O.ai automatically selecting the best suitable predictor model to predict microlithiasis/sludge. The predictor model was further validated in two independent retrospective cohorts from two tertiary care centers (117 patients).

RESULTS

Twenty-eight categorized patients' variables recorded at admission were identified to compute the predictor model with an accuracy of 0.84 [95% confidence interval (CI): 0.791-0.9185], positive predictive value of 0.84, and negative predictive value of 0.80 in the identification cohort (218 patients). In the validation cohort, the robustness of the prediction model was confirmed with an accuracy of 0.76 (95%CI: 0.673-0.8347), positive predictive value of 0.76, and negative predictive value of 0.78 (117 patients).

CONCLUSION

We present a robust and validated ML-based predictor model consisting of routinely recorded parameters at admission that can predict biliary sludge and microlithiasis as the cause of AP.

The role of artificial neural networks in prediction of severe acute pancreatitis associated acute respiratory distress syndrome: A retrospective study

Early identification and intervention of acute respiratory distress syndrome (ARDS) are particularly important. This study aimed to construct predictive models for ARDS following severe acute pancreatitis (SAP) by artificial neural networks and logistic regression. The artificial neural networks model was constructed using clinical data from 214 SAP patients. The patient cohort was randomly divided into a training set and a test set, with 149 patients allocated to the training set and 65 patients assigned to the test set. The artificial neural networks and logistic regression models were trained by the training set, and then the performance of both models was evaluated using the test set. The sensitivity, specificity, PPV, NPV, accuracy, and AUC value of artificial neural networks model were 68.0%, 87.5%, 77.3%, 81.4%, 80.0%, 0.853 ± 0.054 (95% CI: 0.749-0.958). The sensitivity, specificity, PPV, NPV, accuracy and AUC value of logistic regression model were 48.7%, 85.3%, 65.5%, 74.4%, 72.0%, 0.799 ± 0.045 (95% CI: 0.710-0.888). There were no significant differences between the artificial neural networks and logistic regression models in predictive performance. Bedside Index of Severity in Acute Pancreatitis score, procalcitonin, prothrombin time, and serum calcium were the most important predictive variables in the artificial neural networks model. The discrimination abilities of logistic regression and artificial neural networks models in predicting SAP-related ARDS were similar. It is advisable to choose the model according to the specific research purpose.

Comparing Machine Learning and PLSDA Algorithms for Durian Pulp Classification Using Inline NIR Spectra

The aim of this study was to evaluate and compare the performance of multivariate classification algorithms, specifically Partial Least Squares Discriminant Analysis (PLS-DA) and machine learning algorithms, in the classification of Monthong durian pulp based on its dry matter content (DMC) and soluble solid content (SSC), using the inline acquisition of near-infrared (NIR) spectra. A total of 415 durian pulp samples were collected and analyzed. Raw spectra were preprocessed using five different combinations of spectral preprocessing techniques: Moving Average with Standard Normal Variate (MA+SNV), Savitzky-Golay Smoothing with Standard Normal Variate (SG+SNV), Mean Normalization (SG+MN), Baseline Correction (SG+BC), and Multiplicative Scatter Correction (SG+MSC). The results revealed that the SG+SNV preprocessing technique produced the best performance with both the PLS-DA and machine learning algorithms. The optimized wide neural network algorithm of machine learning achieved the highest overall classification accuracy of 85.3%, outperforming the PLS-DA model, with overall classification accuracy of 81.4%. Additionally, evaluation metrics such as recall, precision, specificity, F1-score, AUC ROC, and kappa were calculated and compared between the two models. The findings of this study demonstrate the potential of machine learning algorithms to provide similar or better performance compared to PLS-DA in classifying Monthong durian pulp based on DMC and SSC using NIR spectroscopy, and they can be applied in the quality control and management of durian pulp production and storage.

Early prediction of the severe course, survival, and ICU requirements in acute pancreatitis by artificial intelligence

OBJECTIVE

To evaluate the success of artificial intelligence for early prediction of severe course, survival, and intensive care unit(ICU) requirement in patients with acute pancreatitis(AP).

METHODS

Retrospectively, 1334 patients were included the study. Severity is determined according to the Revised Atlanta Classification(RAC). The success of machine learning(ML) method was evaluated by 13 simple demographic, clinical, etiologic, and laboratory features obtained on ER admission. Additionally, it was evaluated whether Balthazar-computerized tomography severity index(CTSI) at 48-h contributed to success. The dataset was split into two parts, 90% for ML(of which 70% for learning and 30% for testing) and 10% for validation and 5-fold stratified sampling has been utilized. Variable Importance was used in the selection of features during training phase of machine. The Gradient Boost Algorithm trained the machine by KNIME analytics platform. SMOTE has been applied to increase the minority classes for training. The combined effects of the measured features were examined by multivariate logistic regression analysis and reciever operating curve curves of the prediction and confidence of the target variables were obtained.

RESULTS

Accuracy values for the early estimation of Atlanta severity score, ICU requirement, and survival were found as 88.20%, 98.25%, and 92.77% respectively. When Balthazar-CTSI score is used, results were found as 91.02%, 92.25%, and 98% respectively.

CONCLUSIONS

The ML method we used successfully predicted the severe course, ICU requirement and survival, with promising accuracy values of over 88%. If 48-h Balthazar-CTSI is included in the calculation, the severity score and survival rates increase even more.

Big Data in Gastroenterology Research

Studying individual data types in isolation provides only limited and incomplete answers to complex biological questions and particularly falls short in revealing sufficient mechanistic and kinetic details. In contrast, multi-omics approaches to studying health and disease permit the generation and integration of multiple data types on a much larger scale, offering a comprehensive picture of biological and disease processes. Gastroenterology and hepatobiliary research are particularly well-suited to such analyses, given the unique position of the luminal gastrointestinal (GI) tract at the nexus between the gut (mucosa and luminal contents), brain, immune and endocrine systems, and GI microbiome. The generation of 'big data' from multi-omic, multi-site studies can enhance investigations into the connections between these organ systems and organisms and more broadly and accurately appraise the effects of dietary, pharmacological, and other therapeutic interventions. In this review, we describe a variety of useful omics approaches and how they can be integrated to provide a holistic depiction of the human and microbial genetic and proteomic changes underlying physiological and pathophysiological phenomena. We highlight the potential pitfalls and alternatives to help avoid the common errors in study design, execution, and analysis. We focus on the application, integration, and analysis of big data in gastroenterology and hepatobiliary research.

Improving mortality prediction in Acute Pancreatitis by machine learning and data augmentation

Acute Pancreatitis (AP) is the inflammation of the pancreas that can be fatal or lead to further complications based on the severity of the attack. Early detection of AP disease can help save lives by providing utmost care, rigorous treatment, and better resources. In this era of data and technology, instead of relying on manual scoring systems, scientists are employing advanced machine learning and data mining models for the early detection of patients with high chances of mortality. The current work on AP mortality prediction is negligible, and the few studies that exist have many shortcomings and are impractical for clinical deployment. In this research work, we tried to overcome the existing issues. One main issue is the lack of high-quality public datasets for AP, which are crucial for effectively training ML models. The available datasets are small in size, have many missing values, and suffer from high class imbalance. We augmented three public datasets, MIMIC-III, MIMIC-IV, and eICU, to obtain a larger dataset, and experiments proved that augmented data trained classifiers better than original small datasets. Moreover, we employed emerging advanced techniques to handle underlying issues in data. The results showed that iterative imputer is best for filling missing values in AP data. It beats not only the basic techniques but also the Knn-based imputation. Class imbalance is first addressed using data downsampling; apparently, it gave decent results on small test sets. However, we conducted numerous experiments on large test sets to prove that downsampling in the case of AP produced misleading and poor results. Next, we applied various techniques to upsample data in two different class splits, a 50 to 50 and a 70 to 30 majority-minority class split. Four different tabular generative adversarial networks, CTGAN, TGAN, CopulaGAN, and CTAB, and a variational autoencoder, TVAE, were deployed for synthetic data generation. SMOTE was also utilized for data upsampling. The computational results showed that the Random Forest (RF) classifier outperformed all other classifiers on a 50 to 50 class split data generated by CTGAN, with 0.702 Fβ and 0.833 recall. Results produced by RF on the TVAE dataset were also comparable, with 0.698 Fβ. In the case of SMOTE-based upsampling, DNN performed best with a 0.671 Fβ score.

Prediction of the severity of acute pancreatitis using machine learning models

BACKGROUND

Acute pancreatitis (AP) is the most common pancreatic disease. Predicting the severity of AP is critical for making preventive decisions. However, the performance of existing scoring systems in predicting AP severity was not satisfactory. The purpose of this study was to develop predictive models for the severity of AP using machine learning (ML) algorithms and explore the important predictors that affected the prediction results.

METHODS

The data of 441 patients in the Department of Gastroenterology in our hospital were analyzed retrospectively. The demographic data, blood routine and blood biochemical indexes, and the CTSI score were collected to develop five different ML predictive models to predict the severity of AP. The performance of the models was evaluated by the area under the receiver operating characteristic curve (AUC). The important predictors were determined by ranking the feature importance of the predictive factors.

RESULTS

Compared to other ML models, the extreme gradient boosting model (XGBoost) showed better performance in predicting severe AP, with an AUC of 0.906, an accuracy of 0.902, a sensitivity of 0.700, a specificity of 0.961, and a F1socre of 0.764. Further analysis showed that the CTSI score, ALB, LDH, and NEUT were the important predictors of the severity of AP.

CONCLUSION

The results showed that the XGBoost algorithm can accurately predict the severity of AP, which can provide an assistance for the clinicians to identify severe AP at an early stage.

Usefulness of Random Forest Algorithm in Predicting Severe Acute Pancreatitis

BACKGROUND AND AIMS

This study aimed to develop an interpretable random forest model for predicting severe acute pancreatitis (SAP).

METHODS

Clinical and laboratory data of 648 patients with acute pancreatitis were retrospectively reviewed and randomly assigned to the training set and test set in a 3:1 ratio. Univariate analysis was used to select candidate predictors for the SAP. Random forest (RF) and logistic regression (LR) models were developed on the training sample. The prediction models were then applied to the test sample. The performance of the risk models was measured by calculating the area under the receiver operating characteristic (ROC) curves (AUC) and area under precision recall curve. We provide visualized interpretation by using local interpretable model-agnostic explanations (LIME).

RESULTS

The LR model was developed to predict SAP as the following function: -1.10-0.13×albumin (g/L) + 0.016 × serum creatinine (μmol/L) + 0.14 × glucose (mmol/L) + 1.63 × pleural effusion (0/1)(No/Yes). The coefficients of this formula were utilized to build a nomogram. The RF model consists of 16 variables identified by univariate analysis. It was developed and validated by a tenfold cross-validation on the training sample. Variables importance analysis suggested that blood urea nitrogen, serum creatinine, albumin, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, calcium, and glucose were the most important seven predictors of SAP. The AUCs of RF model in tenfold cross-validation of the training set and the test set was 0.89 and 0.96, respectively. Both the area under precision recall curve and the diagnostic accuracy of the RF model were higher than that of both the LR model and the BISAP score. LIME plots were used to explain individualized prediction of the RF model.

CONCLUSIONS

An interpretable RF model exhibited the highest discriminatory performance in predicting SAP. Interpretation with LIME plots could be useful for individualized prediction in a clinical setting. A nomogram consisting of albumin, serum creatinine, glucose, and pleural effusion was useful for prediction of SAP.

Predicting the Need for Therapeutic Intervention and Mortality in Acute Pancreatitis: A Two-Center International Study Using Machine Learning

Background: Current approaches to predicting intervention needs and mortality have reached 65–85% accuracy, which falls below clinical decision-making requirements in patients with acute pancreatitis (AP). We aimed to accurately predict therapeutic intervention needs and mortality on admission, in AP patients, using machine learning (ML). Methods: Data were obtained from three databases of patients admitted with AP: one retrospective (Chengdu) and two prospective (Liverpool and Chengdu) databases. Intervention and mortality differences, as well as potential predictors, were investigated. Univariate analysis was conducted, followed by a random forest ML algorithm used in multivariate analysis, to identify predictors. The ML performance matrix was applied to evaluate the model’s performance. Results: Three datasets of 2846 patients included 25 potential clinical predictors in the univariate analysis. The top ten identified predictors were obtained by ML models, for predicting interventions and mortality, from the training dataset. The prediction of interventions includes death in non-intervention patients, validated with high accuracy (96%/98%), the area under the receiver-operating-characteristic curve (0.90/0.98), and positive likelihood ratios (22.3/69.8), respectively. The post-test probabilities in the test set were 55.4% and 71.6%, respectively, which were considerably superior to existing prognostic scores. The ML model, for predicting mortality in intervention patients, performed better or equally with prognostic scores. Conclusions: ML, using admission clinical predictors, can accurately predict therapeutic interventions and mortality in patients with AP.

Machine learning predictive models for acute pancreatitis: A systematic review

INTRODUCTION

Acute pancreatitis (AP) is a common clinical pancreatic disease. Patients with different severity levels have different clinical outcomes. With the advantages of algorithms, machine learning (ML) has gradually emerged in the field of disease prediction, assisting doctors in decision-making.

METHODS

A systematic review was conducted using the PubMed, Web of Science, Scopus, and Embase databases, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Publication time was limited from inception to 29 May 2021. Studies that have used ML to establish predictive tools for AP were eligible for inclusion. Quality assessment of the included studies was conducted in accordance with the IJMEDI checklist.

RESULTS

In this systematic review, 24 of 2,913 articles, with a total of 8,327 patients and 47 models, were included. The studies could be divided into five categories: 10 studies (42%) reported severity prediction; 10 studies (42%), complication prediction; 3 studies (13%), mortality prediction; 2 studies (8%), recurrence prediction; and 2 studies (8%), surgery timing prediction. ML showed great accuracy in several prediction tasks. However, most of the included studies were retrospective in nature, conducted at a single centre, based on database data, and lacked external validation. According to the IJMEDI checklist and our scoring criteria, two studies were considered to be of high quality. Most studies had an obvious bias in the quality of data preparation, validation, and deployment dimensions.

CONCLUSION

In the prediction tasks for AP, ML has shown great potential in assisting decision-making. However, the existing studies still have some deficiencies in the process of model construction. Future studies need to optimize the deficiencies and further evaluate the comparability of the ML systems and model performance, so as to consequently develop high-quality ML-based models that can be used in clinical practice.