Ensembled deep learning model outperforms human experts in diagnosing biliary atresia from sonographic gallbladder images

Enhanced Macular Telangiectasia Type 2 Detection: Leveraging Self-Supervised Learning and Ensemble Models

Objective

To investigate an ensemble-based approach utilizing deep learning models for accurate and interpretable detection of macular telangiectasia (MacTel) type 2 on OCT imaging.

Design

Retrospective analysis of OCT scans, model development, and assessment.

Participants

A total of 5200 OCT images from participants in the MacTel Registry conducted by the Lowy Medical Research Institute and from the University of Washington (780 MacTel patients and 1900 non-MacTel patients).

Methods Intervention or Testing

We trained multiple individual MacTel vs. non-MacTel classification models using traditional supervised learning and self-supervised learning (SSL) and ensembled them using average weighting methods. We investigated diverse methodologies for constructing the ensemble, including varied architectural configurations and learning paradigms of individual models, and manipulating the amount of labeled data accessible for training. Model performance was compared against human expert graders on held-out test set data. Model interpretability was investigated using gradient-weighted class activation maps (Grad-CAM) visualization and by evaluating interrater agreement.

Main Outcome Measures

For model performance, area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), accuracy, sensitivity, and specificity were reported. For interpretability, interrater agreements and Grad-CAM visualization results were evaluated.

Results

Despite access to only 419 OCT volumes, including 185 MacTel patients within the 10% labeled training dataset, the ensemble model demonstrated a performance level (AUROC 0.972 [95% confidence interval (CI), 0.971-0.973], AUPRC 0.967 [95% CI, 0.965-0.969], accuracy 91.7%, sensitivity 0.905, and specificity 0.925) comparable to the human experts ensemble (AUROC 0.977 [95% CI, 0.975-0.978], AUPRC 0.987 [95% CI, 0.986-0.987], accuracy 96.8%, sensitivity 0.929, and specificity 1) on a test set of 500 patients. The individual models did not achieve the same performance levels when evaluated separately.

Conclusions

Even with limited data, combining SSL with ensemble approaches improved MacTel classification accuracy and interpretation compared to the individual models. Self-supervised learning captures meaningful representations from unlabeled data, a key benefit in the setting of limited data such as with rare diseases.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

MMRNet: Ensemble deep learning models for predicting mismatch repair deficiency in endometrial cancer from histopathological images

Combining molecular classification with clinicopathologic methods improves risk assessment and chooses therapies for endometrial cancer (EC). Detecting mismatch repair (MMR) deficiencies in EC is crucial for screening Lynch syndrome and identifying immunotherapy candidates. An affordable and accessible tool is urgently needed to determine MMR status in EC patients. We introduce MMRNet, a deep convolutional neural network designed to predict MMR-deficient EC from whole-slide images stained with hematoxylin and eosin. MMRNet demonstrates strong performance, achieving an average area under the receiver operating characteristic curve (AUROC) of 0.897, with a sensitivity of 0.628 and a specificity of 0.949 in internal cross-validation. External validation using three additional datasets results in AUROCs of 0.790, 0.807, and 0.863. Employing a human-machine fusion approach notably improves diagnostic accuracy. MMRNet presents an effective method for identifying EC cases for confirmatory MMR testing and may assist in selecting candidates for immunotherapy.

Beyond human perception: challenges in AI interpretability of orangutan artwork

Drawings serve as a profound medium of expression for both humans and apes, offering unique insights into the cognitive and emotional landscapes of the artists, regardless of their species. This study employs artificial intelligence (AI), specifically Convolutional Neural Networks (CNNs) and the interpretability tool Captum, to analyse non-figurative drawings by Molly, an orangutan. The research utilizes VGG19 and ResNet18 models to decode seasonal nuances in the drawings, achieving notable accuracy in seasonal classification and revealing complex influences beyond human-centric methods. Techniques, such as occlusion, integrated gradients, PCA, t-SNE, and Louvain clustering, highlight critical areas and elements influencing seasonal recognition, providing deeper insights into the drawings. This approach not only advances the analysis of non-human art but also demonstrates the potential of AI to enrich our understanding of non-human cognitive and emotional expressions, with significant implications for fields like evolutionary anthropology and comparative psychology.

Transfer learning method for prenatal ultrasound diagnosis of biliary atresia

Biliary atresia (BA) is a rare and severe congenital disorder with a significant challenge for prenatal diagnosis. This study, registered at the Chinese Clinical Trial Registry (ChiCTR2200059705), aimed to develop an intelligent model to aid in the prenatal diagnosis of BA. To develop and evaluate this model, fetuses from 20 hospitals across China and infants sourced from public database were collected. The transfer-learning model (TLM) demonstrated superior diagnostic performance compared to the basic deep-learning model, with higher area under the curves of 0.906 (95%CI: 0.872-0.940) vs 0.793 (0.743-0.843), 0.914 (0.875-0.953) vs 0.790 (0.727-0.853), and 0.907 (0.869-0.945) vs 0.880 (0.838-0.922) for the three independent test cohorts. Furthermore, when aided by the TLM, diagnostic accuracy surpassed that of individual sonologists alone. The TLM achieved satisfactory performance in predicting fetal BA, providing a low-cost, easily accessible, and accurate diagnostic tool for this condition, making it an effective aid in clinical practice.

Development of an artificial intelligence-based multimodal diagnostic system for early detection of biliary atresia

BACKGROUND

Early diagnosis of biliary atresia (BA) is crucial for improving patient outcomes, yet remains a significant global challenge. This challenge may be ameliorated through the application of artificial intelligence (AI). Despite the promise of AI in medical diagnostics, its application to multimodal BA data has not yet achieved substantial breakthroughs. This study aims to leverage diverse data sources and formats to develop an intelligent diagnostic system for BA.

METHODS

We constructed the largest known multimodal BA dataset, comprising ultrasound images, clinical data, and laboratory results. Using this dataset, we developed a novel deep learning model and simplified it using easily obtainable data, eliminating the need for blood samples. The models were externally validated in a prospective study. We compared the performance of our model with human experts of varying expertise levels and evaluated the AI system's potential to enhance its diagnostic accuracy.

RESULTS

The retrospective study included 1579 participants. The multimodal model achieved an AUC of 0.9870 on the internal test set, outperforming human experts. The simplified model yielded an AUC of 0.9799. In the prospective study's external test set of 171 cases, the multimodal model achieved an AUC of 0.9740, comparable to that of a radiologist with over 10 years of experience (AUC = 0.9766). For less experienced radiologists, the AI-assisted diagnostic AUC improved from 0.6667 to 0.9006.

CONCLUSIONS

This AI-based screening application effectively facilitates early diagnosis of BA and serves as a valuable reference for addressing common challenges in rare diseases. The model's high accuracy and its ability to enhance the diagnostic performance of human experts underscore its potential for significant clinical impact.

A deep learning based smartphone application for early detection of nasopharyngeal carcinoma using endoscopic images

Nasal endoscopy is crucial for the early detection of nasopharyngeal carcinoma (NPC), but its accuracy relies heavily on the clinician's expertise, posing challenges for primary healthcare providers. Here, we retrospectively analysed 39,340 nasal endoscopic white-light images from three high-incidence NPC centres, utilising eight advanced deep learning models to develop an Internet-enabled smartphone application, "Nose-Keeper", that can be used for early detection of NPC and five prevalent nasal diseases and assessment of healthy individuals. Our app demonstrated a remarkable overall accuracy of 92.27% (95% Confidence Interval (CI): 90.66%-93.61%). Notably, its sensitivity and specificity in NPC detection achieved 96.39% and 99.91%, respectively, outperforming nine experienced otolaryngologists. Explainable artificial intelligence was employed to highlight key lesion areas, improving Nose-Keeper's decision-making accuracy and safety. Nose-Keeper can assist primary healthcare providers in diagnosing NPC and common nasal diseases efficiently, offering a valuable resource for people in high-incidence NPC regions to manage nasal cavity health effectively.

Inclusive and Accurate Clinical Diagnostics Using Intelligent Computation and Smartphone Imaging

Smartphone-based colorimetry has been widely applied in clinical analysis, although significant challenges remain in its practical implementation, including the need to consider biases introduced by the ambient imaging environment, which limit its potential within a clinical decision pathway. In addition, most commercial devices demonstrate variability introduced by manufacturer-to-manufacturer differences. Here, we undertake a systematic characterization of the potential imaging interferences that lead to this limited performance in conventional smartphones and, in doing so, provide a comprehensive new understanding of smartphone color imaging. Through derivation of a strongly correlated parameter for sample quantification, we enable real-time imaging, which for the first time, takes the first steps to turning the mobile phone camera into an analytical instrument - irrespective of model, software, and the operating systems used. We demonstrate clinical applicability through the imaging of patients' skin, enabling rapid and convenient diagnosis of cyanosis and measurement of local oxygen concentration to a level that unlocks clinical decision-making for monitoring cardiovascular disease and anemia. Importantly, we show that our solution also accounts for the differences in individuals' skin tones as measured across the Fitzpatrick scale, overcoming potential clinically significant errors in current optical oximetry.

Deep learning-driven ultrasound-assisted diagnosis: optimizing GallScopeNet for precise identification of biliary atresia

Background

Biliary atresia (BA) is a severe congenital biliary developmental abnormality threatening neonatal health. Traditional diagnostic methods rely heavily on experienced radiologists, making the process time-consuming and prone to variability. The application of deep learning for the automated diagnosis of BA remains underexplored.

Methods

This study introduces GallScopeNet, a deep learning model designed to improve diagnostic efficiency and accuracy through innovative architecture and advanced feature extraction techniques. The model utilizes data from a carefully constructed dataset of gallbladder ultrasound images. A dataset comprising thousands of ultrasound images was employed, with the majority used for training and validation and a subset reserved for external testing. The model's performance was evaluated using five-fold cross-validation and external assessment, employing metrics such as accuracy and the area under the receiver operating characteristic curve (AUC), compared against clinical diagnostic standards.

Results

GallScopeNet demonstrated exceptional performance in distinguishing BA from non-BA cases. In the external test dataset, GallScopeNet achieved an accuracy of 81.21% and an AUC of 0.85, indicating strong diagnostic capabilities. The results highlighted the model's ability to maintain high classification performance, reducing misdiagnosis and missed diagnosis.

Conclusion

GallScopeNet effectively differentiates between BA and non-BA images, demonstrating significant potential and reliability for early diagnosis. The system's high efficiency and accuracy suggest it could serve as a valuable diagnostic tool in clinical settings, providing substantial technical support for improving diagnostic workflows.

Fully automated segmentation and volumetric measurement of ocular adnexal lymphoma by deep learning-based self-configuring nnU-net on multi-sequence MRI: a multi-center study

PURPOSE

To evaluate nnU-net's performance in automatically segmenting and volumetrically measuring ocular adnexal lymphoma (OAL) on multi-sequence MRI.

METHODS

We collected T1-weighted (T1), T2-weighted and T1-weighted contrast-enhanced images with/without fat saturation (T2_FS/T2_nFS, T1c_FS/T1c_nFS) of OAL from four institutions. Two radiologists manually annotated lesions as the ground truth using ITK-SNAP. A deep learning framework, nnU-net, was developed and trained using two models. Model 1 was trained on T1, T2, and T1c, while Model 2 was trained exclusively on T1 and T2. A 5-fold cross-validation was utilized in the training process. Segmentation performance was evaluated using the Dice similarity coefficient (DSC), sensitivity, and positive prediction value (PPV). Volumetric assessment was performed using Bland-Altman plots and Lin's concordance correlation coefficient (CCC).

RESULTS

A total of 147 patients from one center were selected as training set and 33 patients from three centers were regarded as test set. For both Model 1 and 2, nnU-net demonstrated outstanding segmentation performance on T2_FS with DSC of 0.80-0.82, PPV of 84.5-86.1%, and sensitivity of 77.6-81.2%, respectively. Model 2 failed to detect 19 cases of T1c, whereas the DSC, PPV, and sensitivity for T1_nFS were 0.59, 91.2%, and 51.4%, respectively. Bland-Altman plots revealed minor tumor volume differences with 0.22-1.24 cm3 between nnU-net prediction and ground truth on T2_FS. The CCC were 0.96 and 0.93 in Model 1 and 2 for T2_FS images, respectively.

CONCLUSION

The nnU-net offered excellent performance in automated segmentation and volumetric assessment in MRI of OAL, particularly on T2_FS images.

Automated gall bladder cancer detection using artificial gorilla troops optimizer with transfer learning on ultrasound images

The gallbladder (GB) is a small pouch and a deep tissue placed under the liver. GB Cancer (GBC) is a deadly illness that is complex to discover in an initial phase. Initial diagnosis can significantly enhance the existence rate. Non-ionizing energy, low cost, and convenience make the US a general non-invasive analytical modality for patients with GB diseases. Automatic recognition of GBC from US imagery is a significant issue that has gained much attention from researchers. Recently, machine learning (ML) techniques dependent on convolutional neural network (CNN) architectures have prepared transformational growth in radiology and medical analysis for illnesses like lung, pancreatic, breast, and melanoma. Deep learning (DL) is a region of artificial intelligence (AI), a functional medical tomography model that can help in the initial analysis of GBC. This manuscript presents an Automated Gall Bladder Cancer Detection using an Artificial Gorilla Troops Optimizer with Transfer Learning (GBCD-AGTOTL) technique on Ultrasound Images. The GBCD-AGTOTL technique examines the US images for the presence of gall bladder cancer using the DL model. In the initial stage, the GBCD-AGTOTL technique preprocesses the US images using a median filtering (MF) approach. The GBCD-AGTOTL technique applies the Inception module for feature extraction, which learns the complex and intrinsic patterns in the pre-processed image. Besides, the AGTO algorithm-based hyperparameter tuning procedure takes place, which optimally picks the hyperparameter values of the Inception technique. Lastly, the bidirectional gated recurrent unit (BiGRU) model helps classify gall bladder cancer. A series of simulation analyses were performed to ensure the performance of the GBCD-AGTOTL technique on the GBC dataset. The experimental outcomes inferred the enhanced abilities of the GBCD-AGTOTL in detecting gall bladder cancer.

Prediction of Parkinson's disease by transcranial sonography-based deep learning

OBJECTIVES

Transcranial sonography has been used as a valid neuroimaging tool to diagnose Parkinson's disease (PD). This study aimed to develop a modified transcranial sonography (TCS) technique based on a deep convolutional neural network (DCNN) model to predict Parkinson's disease.

METHODS

This retrospective diagnostic study was conducted using 1529 transcranial sonography images collected from 854 patients with PD and 775 normal controls admitted to the Second Affiliated Hospital of Soochow University (Suzhou, Jiangsu, China) between September 2019 and May 2022. The data set was divided into training cohorts (570 PD patients and 541 normal controls), and the validation set (184 PD patients and 234 normal controls). Using these datasets, we developed four different DCNN models (ResNet18, ResNet50, ResNet152, and DenseNet121). We then assessed their diagnostic performance, including the area under the receiver operating characteristic (AUROC) curve, specificity, sensitivity, positive predictive value (PPV), negative predictive value (NPV), and F1 score and compared with traditional diagnostic criteria.

RESULTS

Among the 1529 TCS images, 570 PD patients and 541 normal controls from 4 of 6 sonographers of the TCS team were selected as the training cohort, and 184 PD patients and 234 normal controls from the other 2 sonographers were chosen as the validation cohort. There were no sex and age differences between PD patients and normal control subjects in the training and validation cohorts (P values > 0.05). All DCNN models achieved good performance in distinguishing PD patients from normal control subjects on the validation datasets, with diagnostic AUROCs and accuracy of 0.949 (95% CI 0.925, 0.965) and 86.60 for the RestNet18 model, 0.949 (95% CI 0.929, 0.971) and 87.56 for ResNet50, 0.945 (95% CI 0.931, 0.969) and 88.04 for ResNet152, 0.953 (95% CI 0.935, 0.971) and 87.80 for DenseNet121, respectively. On the other hand, the diagnostic accuracy of the traditional diagnostic method was 82.30. The accuracy of all DCNN models was higher than that of traditional diagnostic method. Moreover, the 5k-fold cross-validation results in train datasets showed that these DCNN models are robust.

CONCLUSION

The developed transcranial sonography-based DCNN models performed better than traditional diagnostic criteria, thus improving the sonographer's accuracy in diagnosing PD.

Comparison for the diagnostic performance of early diagnostic methods for biliary atresia: a systematic review and network meta-analysis

BACKGROUND

Biliary atresia (BA), a progressive condition affecting canalicular-bile duct function/anatomy, requires prompt surgical intervention for favorable outcomes. Therefore, we conducted a network meta-analysis of common diagnostic methods to assess their performance and provide evidence-based support for clinical decision-making.

METHODS

We reviewed literature in PubMed, EMBASE, and Cochrane for BA diagnostics. The search included gamma-glutamyl transferase (GGT), direct/combined bilirubin, matrix metalloproteinase 7 (MMP-7), ultrasonic triangular cord sign (TCS), hepatic scintigraphy (HS), and percutaneous cholangiocholangiography/percutaneous transhepatic cholecysto-cholangiography (PCC/PTCC). QUADAS-2 assessed study quality. Heterogeneity and threshold effect were evaluated using I2 and Spearman's correlation. We combined effect estimates, constructed SROC models, and conducted a network meta-analysis based on the ANOVA model, along with meta-regression and subgroup analysis, to obtain precise diagnostic performance assessments for BA.

RESULTS

A total of 40 studies were included in our analysis. GGT demonstrated high diagnostic accuracy for BA with a sensitivity of 81.5% (95% CI 0.792-0.836) and specificity of 72.1% (95% CI 0.693-0.748). Direct bilirubin/conjugated bilirubin showed a sensitivity of 87.6% (95% CI 0.833-0.911) but lower specificity of 59.4% (95% CI 0.549-0.638). MMP-7 exhibited a total sensitivity of 91.5% (95% CI 0.893-0.934) and a specificity of 84.3% (95% CI 0.820-0.863). TCS exhibited a sensitivity of 58.1% (95% CI 0.549-0.613) and high specificity of 92.9% (95% CI 0.911-0.944). HS had a high sensitivity of 98.4% (95% CI 0.968-0.994) and moderate specificity of 79.0% (95% CI 0.762-0.816). PCC/PTCC exhibited excellent diagnostic performance with a sensitivity of 100% (95% CI 0.900-1.000) and specificity of 87.0% (95% CI 0.767-0.939). Based on the ANOVA model, the network meta-analysis revealed that MMP-7 ranked second overall, with PCC/PTCC ranking first, both exhibiting superior diagnostic accuracy compared to other techniques. Our analysis showed no significant bias in most methodologies, but MMP-7 and hepatobiliary scintigraphy exhibited biases, with p values of 0.023 and 0.002, respectively.

CONCLUSION

MMP-7 and ultrasound-guided PCC/PTCC show diagnostic potential in the early diagnosis of BA, but their clinical application is restricted due to practical limitations. Currently, the cutoff value of MMP-7 is unclear, and further evidence-based medical research is needed to firmly establish its diagnostic value. Until more evidence is available, MMP-7 is not suitable for widespread diagnostic use. Therefore, considering cost and operational simplicity, liver function tests combined with ultrasound remain the most clinically valuable non-invasive diagnostic methods for BA.

Predictive modeling for early detection of biliary atresia in infants with cholestasis: Insights from a machine learning study

Cholestasis, characterized by the obstruction of bile flow, poses a significant concern in neonates and infants. It can result in jaundice, inadequate weight gain, and liver dysfunction. However, distinguishing between biliary atresia (BA) and non-biliary atresia in these young patients presenting with cholestasis poses a formidable challenge, given the similarity in their clinical manifestations. To this end, our study endeavors to construct a screening model aimed at prognosticating outcomes in cases of BA. Within this study, we introduce a wrapper feature selection model denoted as bWFMVO-SVM-FS, which amalgamates the water flow-based multi-verse optimizer (WFMVO) and support vector machine (SVM) technology. Initially, WFMVO is benchmarked against eleven state-of-the-art algorithms, with its efficiency in searching for optimized feature subsets within the model validated on IEEE CEC 2017 and IEEE CEC 2022 benchmark functions. Subsequently, the developed bWFMVO-SVM-FS model is employed to analyze a cohort of 870 consecutively registered cases of neonates and infants with cholestasis (diagnosed as either BA or non-BA) from Xinhua Hospital and Shanghai Children's Hospital, both affiliated with Shanghai Jiao Tong University. The results underscore the remarkable predictive capacity of the model, achieving an accuracy of 92.639 % and specificity of 88.865 %. Gamma-glutamyl transferase, triangular cord sign, weight, abnormal gallbladder, and stool color emerge as highly correlated with early symptoms in BA infants. Furthermore, leveraging these five significant features enhances the interpretability of the machine learning model's performance outcomes for medical professionals, thereby facilitating more effective clinical decision-making.

Combining quantitative and qualitative analysis for scoring pleural line in lung ultrasound

Objective.Accurate assessment of pleural line is crucial for the application of lung ultrasound (LUS) in monitoring lung diseases, thereby aim of this study is to develop a quantitative and qualitative analysis method for pleural line.Approach.The novel cascaded deep learning model based on convolution and multilayer perceptron was proposed to locate and segment the pleural line in LUS images, whose results were applied for quantitative analysis of textural and morphological features, respectively. By using gray-level co-occurrence matrix and self-designed statistical methods, eight textural and three morphological features were generated to characterize the pleural lines. Furthermore, the machine learning-based classifiers were employed to qualitatively evaluate the lesion degree of pleural line in LUS images.Main results.We prospectively evaluated 3770 LUS images acquired from 31 pneumonia patients. Experimental results demonstrated that the proposed pleural line extraction and evaluation methods all have good performance, with dice and accuracy of 0.87 and 94.47%, respectively, and the comparison with previous methods found statistical significance (P< 0.001 for all). Meanwhile, the generalization verification proved the feasibility of the proposed method in multiple data scenarios.Significance.The proposed method has great application potential for assessment of pleural line in LUS images and aiding lung disease diagnosis and treatment.

Diagnosis of Gallbladder Disease Using Artificial Intelligence: A Comparative Study

Gallbladder (GB) disease is a common pathology that needs correct and early diagnosis for the optimum medical treatment. Early diagnosis is crucial as any delay or misdiagnosis can worsen the patient situation. Incorrect diagnosis could also lead to an escalation in patient symptoms and poorer clinical outcomes. The use of Artificial Intelligence (AI) techniques, ranging from Machine Learning (ML) to Deep Learning (DL) to predict disease progression, identify abnormalities, and estimate mortality rates associated with GB disorders has increased over the past decade. To this end, this paper provides a comprehensive overview of the AI approaches used in the diagnosis of GB illnesses. This review compiles and compares relevant papers from the last decade to show how AI might enhance diagnostic precision, speed, and efficiency. Therefore, this survey gives researchers the opportunity to find out both the diagnosis of GB diseases and AI techniques in one place. The maximum accuracy rate by ML was when using SVM with 96.67%, whilst the maximum accuracy rate by DL was by utilising a unique structure of VGG, GoogleNet, ResNet, AlexNet and Inception with 98.77%. This could provide a clear path for further investigations and algorithm’s development to boost diagnostic results to improve the patient’s condition and choose the appropriate treatment.

Interpretable artificial intelligence-based app assists inexperienced radiologists in diagnosing biliary atresia from sonographic gallbladder images

BACKGROUND

A previously trained deep learning-based smartphone app provides an artificial intelligence solution to help diagnose biliary atresia from sonographic gallbladder images, but it might be impractical to launch it in real clinical settings. This study aimed to redevelop a new model using original sonographic images and their derived smartphone photos and then test the new model's performance in assisting radiologists with different experiences to detect biliary atresia in real-world mimic settings.

METHODS

A new model was first trained retrospectively using 3659 original sonographic gallbladder images and their derived 51,226 smartphone photos and tested on 11,410 external validation smartphone photos. Afterward, the new model was tested in 333 prospectively collected sonographic gallbladder videos from 207 infants by 14 inexperienced radiologists (9 juniors and 5 seniors) and 4 experienced pediatric radiologists in real-world mimic settings. Diagnostic performance was expressed as the area under the receiver operating characteristic curve (AUC).

RESULTS

The new model outperformed the previously published model in diagnosing BA on the external validation set (AUC 0.924 vs 0.908, P = 0.004) with higher consistency (kappa value 0.708 vs 0.609). When tested in real-world mimic settings using 333 sonographic gallbladder videos, the new model performed comparable to experienced pediatric radiologists (average AUC 0.860 vs 0.876) and outperformed junior radiologists (average AUC 0.838 vs 0.773) and senior radiologists (average AUC 0.829 vs 0.749). Furthermore, the new model could aid both junior and senior radiologists to improve their diagnostic performances, with the average AUC increasing from 0.773 to 0.835 for junior radiologists and from 0.749 to 0.805 for senior radiologists.

CONCLUSIONS

The interpretable app-based model showed robust and satisfactory performance in diagnosing biliary atresia, and it could aid radiologists with limited experiences to improve their diagnostic performances in real-world mimic settings.

Spectroscopy-Guided Deep Learning Predicts Solid-Liquid Surface Adsorbate Properties in Unseen Solvents

Accurately and rapidly acquiring the microscopic properties of a material is crucial for catalysis and electrochemistry. Characterization tools, such as spectroscopy, can be a valuable tool to infer these properties, and when combined with machine learning tools, they can theoretically achieve fast and accurate prediction results. However, on the path to practical applications, training a reliable machine learning model is faced with the challenge of uneven data distribution in a vast array of non-negligible solvent types. Herein, we employ a combination of the first-principles-based approach and data-driven model. Specifically, we utilize density functional theory (DFT) to calculate theoretical spectral data of CO-Ag adsorption in 23 different solvent systems as a data source. Subsequently, we propose a hierarchical knowledge extraction multiexpert neural network (HMNN) to bridge the knowledge gaps among different solvent systems. HMNN undergoes two training tiers: in tier I, it learns fundamental quantitative spectra-property relationships (QSPRs), and in tier II, it inherits the fundamental QSPR knowledge from previous steps through a dynamic integration of expert modules and subsequently captures the solvent differences. The results demonstrate HMNN's superiority in estimating a range of molecular adsorbate properties, with an error range of less than 0.008 eV for zero-shot predictions on unseen solvents. The findings underscore the usability, reliability, and convenience of HMNN and could pave the way for real-time access to microscopic properties by exploiting QSPR.

Is the diagnostic model based on convolutional neural network superior to pediatric radiologists in the ultrasonic diagnosis of biliary atresia?

Background

Many screening and diagnostic methods are currently available for biliary atresia (BA), but the early and accurate diagnosis of BA remains a challenge with existing methods. This study aimed to use deep learning algorithms to intelligently analyze the ultrasound image data, build a BA ultrasound intelligent diagnostic model based on the convolutional neural network, and realize an intelligent diagnosis of BA.

Methods

A total of 4,887 gallbladder ultrasound images of infants with BA, non-BA hyperbilirubinemia, and healthy infants were collected. Two mask region convolutional neural network (Mask R-CNN) models based on different backbone feature extraction networks were constructed. The diagnostic performance between the two models was compared through good-quality images at the image level and the patient level. The diagnostic performance between the two models was compared through poor-quality images. The diagnostic performance of BA between the model and four pediatric radiologists was compared at the image level and the patient level.

Results

The classification performance of BA in model 2 was slightly higher than that in model 1 in the test set, both at the image level and at the patient level, with a significant difference of p = 0.0365 and p = 0.0459, respectively. The classification accuracy of model 2 was slightly higher than that of model 1 in poor-quality images (88.3% vs. 86.4%), and the difference was not statistically significant (p = 0.560). The diagnostic performance of model 2 was similar to that of the two radiology experts at the image level, and the differences were not statistically significant. The diagnostic performance of model 2 in the test set was higher than that of the two radiology experts at the patient level (all p < 0.05).

Conclusion

The performance of model 2 based on Mask R-CNN in the diagnosis of BA reached or even exceeded the level of pediatric radiology experts.

AI-driven decision support systems and epistemic reliance: a qualitative study on obstetricians' and midwives' perspectives on integrating AI-driven CTG into clinical decision making

BACKGROUND

Given that AI-driven decision support systems (AI-DSS) are intended to assist in medical decision making, it is essential that clinicians are willing to incorporate AI-DSS into their practice. This study takes as a case study the use of AI-driven cardiotography (CTG), a type of AI-DSS, in the context of intrapartum care. Focusing on the perspectives of obstetricians and midwives regarding the ethical and trust-related issues of incorporating AI-driven tools in their practice, this paper explores the conditions that AI-driven CTG must fulfill for clinicians to feel justified in incorporating this assistive technology into their decision-making processes regarding interventions in labor.

METHODS

This study is based on semi-structured interviews conducted online with eight obstetricians and five midwives based in England. Participants were asked about their current decision-making processes about when to intervene in labor, how AI-driven CTG might enhance or disrupt this process, and what it would take for them to trust this kind of technology. Interviews were transcribed verbatim and analyzed with thematic analysis. NVivo software was used to organize thematic codes that recurred in interviews to identify the issues that mattered most to participants. Topics and themes that were repeated across interviews were identified to form the basis of the analysis and conclusions of this paper.

RESULTS

There were four major themes that emerged from our interviews with obstetricians and midwives regarding the conditions that AI-driven CTG must fulfill: (1) the importance of accurate and efficient risk assessments; (2) the capacity for personalization and individualized medicine; (3) the lack of significance regarding the type of institution that develops technology; and (4) the need for transparency in the development process.

CONCLUSIONS

Accuracy, efficiency, personalization abilities, transparency, and clear evidence that it can improve outcomes are conditions that clinicians deem necessary for AI-DSS to meet in order to be considered reliable and therefore worthy of being incorporated into the decision-making process. Importantly, healthcare professionals considered themselves as the epistemic authorities in the clinical context and the bearers of responsibility for delivering appropriate care. Therefore, what mattered to them was being able to evaluate the reliability of AI-DSS on their own terms, and have confidence in implementing them in their practice.

Deep learning nomogram for preoperative distinction between Xanthogranulomatous cholecystitis and gallbladder carcinoma: A novel approach for surgical decision

The distinction between Xanthogranulomatous Cholecystitis (XGC) and Gallbladder Carcinoma (GBC) is challenging due to their similar imaging features. This study aimed to differentiate between XGC and GBC using a deep learning nomogram model built from contrast enhanced computed tomography (CT) scans. 297 patients were included with confirmed XGC (94) and GBC (203) as the training and internal validation cohort from 2017 to 2021. The deep learning model Resnet-18 with Fourier transformation named FCovResnet18, shows most impressive potential in distinguishing XGC from GBC using 3-phase merged images. The accuracy, precision and area under the curve (AUC) of the model were then calculated. An additional cohort of 74 patients consisting of 22 XGC and 52 GBC patients was enrolled from two subsidiary hospitals as the external validation cohort. The accuracy, precision and AUC achieve 0.98, 0.99, 1.00 in the internal validation cohort and 0.89, 0.92, 0.92 in external validation cohort. A nomogram model combining clinical characteristics and deep learning prediction score showed improved predicting value. Altogether, FCovResnet18 nomogram has demonstrated its ability to effectively differentiate XGC from GBC preoperatively, which significantly aid surgeons in making informed and accurate surgical decisions for XGC and GBC patients.

Risk stratification of gallbladder masses by machine learning-based ultrasound radiomics models: a prospective and multi-institutional study

OBJECTIVE

This study aimed to evaluate the diagnostic performance of machine learning (ML)-based ultrasound (US) radiomics models for risk stratification of gallbladder (GB) masses.

METHODS

We prospectively examined 640 pathologically confirmed GB masses obtained from 640 patients between August 2019 and October 2022 at four institutions. Radiomics features were extracted from grayscale US images and germane features were selected. Subsequently, 11 ML algorithms were separately used with the selected features to construct optimum US radiomics models for risk stratification of the GB masses. Furthermore, we compared the diagnostic performance of these models with the conventional US and contrast-enhanced US (CEUS) models.

RESULTS

The optimal XGBoost-based US radiomics model for discriminating neoplastic from non-neoplastic GB lesions showed higher diagnostic performance in terms of areas under the curves (AUCs) than the conventional US model (0.822-0.853 vs. 0.642-0.706, p < 0.05) and potentially decreased unnecessary cholecystectomy rate in a speculative comparison with performing cholecystectomy for lesions sized over 10 mm (2.7-13.8% vs. 53.6-64.9%, p < 0.05) in the validation and test sets. The AUCs of the XGBoost-based US radiomics model for discriminating carcinomas from benign GB lesions were higher than the conventional US model (0.904-0.979 vs. 0.706-0.766, p < 0.05). The XGBoost-US radiomics model performed better than the CEUS model in discriminating GB carcinomas (AUC: 0.995 vs. 0.902, p = 0.011).

CONCLUSIONS

The proposed ML-based US radiomics models possess the potential capacity for risk stratification of GB masses and may reduce the unnecessary cholecystectomy rate and use of CEUS.

CLINICAL RELEVANCE STATEMENT

The machine learning-based ultrasound radiomics models have potential for risk stratification of gallbladder masses and may potentially reduce unnecessary cholecystectomies.

KEY POINTS

• The XGBoost-based US radiomics models are useful for the risk stratification of GB masses. • The XGBoost-based US radiomics model is superior to the conventional US model for discriminating neoplastic from non-neoplastic GB lesions and may potentially decrease unnecessary cholecystectomy rate for lesions sized over 10 mm in comparison with the current consensus guideline. • The XGBoost-based US radiomics model could overmatch CEUS model in discriminating GB carcinomas from benign GB lesions.

Prediction of sentinel lymph node metastasis in breast cancer by using deep learning radiomics based on ultrasound images

Sentinel lymph node metastasis (SLNM) is a crucial predictor for breast cancer treatment and survival. This study was designed to propose deep learning (DL) models based on grayscale ultrasound, color Doppler flow imaging (CDFI), and elastography images, and to evaluate how DL radiomics can be used to classify SLNM in breast cancer. Clinical and ultrasound data of 317 patients diagnosed with breast cancer at the Second Affiliated Hospital of Nanchang University were collected from January 2018 to December 2021 and randomly divided into training and internal validation cohorts at a ratio of 7:3. An external validation cohort comprising data from Nanchang Third Hospital with 42 patients collected. Three DL models, namely DL-grayscale, DL-CDFI, and DL-elastography, were proposed to predict SLNM by analyzing grayscale ultrasound, CDFI, and elastography images. Three DL models were compared and evaluated to assess diagnostic performance based on the area under the curve (AUC). The AUCs of the DL-grayscale were 0.855 and 0.788 in the internal and external validation cohorts, respectively. For the DL-CDFI model, the AUCs were 0.761 and 0.728, respectively. The diagnostic performance of DL-elastography was superior to that of the DL-grayscale and DL-CDFI. The AUC of the DL-elastography model was 0.879 in the internal validation cohort, with a classification accuracy of 86.13%, sensitivity of 91.60%, and specificity of 82.79%. The generalization capability of DL-elastography remained high in the external cohort, with an AUC of 0.876, and an accuracy of 85.00%. DL radiomics can be used to classify SLNM in breast cancer using ultrasound images. The proposed DL-elastography model based on elastography images achieved the best diagnostic performance and holds good potential for the management of patients with SLNM.

Artificial intelligence-based prediction of cervical lymph node metastasis in papillary thyroid cancer with CT

OBJECTIVES

To develop an artificial intelligence (AI) system for predicting cervical lymph node metastasis (CLNM) preoperatively in patients with papillary thyroid cancer (PTC) based on CT images.

METHODS

This multicenter retrospective study included the preoperative CT of PTC patients who were divided into the development, internal, and external test sets. The region of interest of the primary tumor was outlined manually on the CT images by a radiologist who has eight years of experience. With the use of the CT images and lesions masks, the deep learning (DL) signature was developed by the DenseNet combined with convolutional block attention module. One-way analysis of variance and least absolute shrinkage and selection operator were used to select features, and a support vector machine was used to construct the radiomics signature. Random forest was used to combine the DL, radiomics, and clinical signature to perform the final prediction. The receiver operating characteristic curve, sensitivity, specificity, and accuracy were used by two radiologists (R1 and R2) to evaluate and compare the AI system.

RESULTS

For the internal and external test set, the AI system achieved excellent performance with AUCs of 0.84 and 0.81, higher than the DL (p = .03, .82), radiomics (p < .001, .04), and clinical model (p < .001, .006). With the aid of the AI system, the specificities of radiologists were improved by 9% and 15% for R1 and 13% and 9% for R2, respectively.

CONCLUSIONS

The AI system can help predict CLNM in patients with PTC, and the radiologists' performance improved with AI assistance.

CLINICAL RELEVANCE STATEMENT

This study developed an AI system for preoperative prediction of CLNM in PTC patients based on CT images, and the radiologists' performance improved with AI assistance, which could improve the effectiveness of individual clinical decision-making.

KEY POINTS

• This multicenter retrospective study showed that the preoperative CT image-based AI system has the potential for predicting the CLNM of PTC. • The AI system was superior to the radiomics and clinical model in predicting the CLNM of PTC. • The radiologists' diagnostic performance improved when they received the AI system assistance.

GIT-Net: An Ensemble Deep Learning-Based GI Tract Classification of Endoscopic Images

This paper presents an ensemble of pre-trained models for the accurate classification of endoscopic images associated with Gastrointestinal (GI) diseases and illnesses. In this paper, we propose a weighted average ensemble model called GIT-NET to classify GI-tract diseases. We evaluated the model on a KVASIR v2 dataset with eight classes. When individual models are used for classification, they are often prone to misclassification since they may not be able to learn the characteristics of all the classes adequately. This is due to the fact that each model may learn the characteristics of specific classes more efficiently than the other classes. We propose an ensemble model that leverages the predictions of three pre-trained models, DenseNet201, InceptionV3, and ResNet50 with accuracies of 94.54%, 88.38%, and 90.58%, respectively. The predictions of the base learners are combined using two methods: model averaging and weighted averaging. The performances of the models are evaluated, and the model averaging ensemble has an accuracy of 92.96% whereas the weighted average ensemble has an accuracy of 95.00%. The weighted average ensemble outperforms the model average ensemble and all individual models. The results from the evaluation demonstrate that utilizing an ensemble of base learners can successfully classify features that were incorrectly learned by individual base learners.

Detection of Gallbladder Disease Types Using Deep Learning: An Informative Medical Method

Nowadays, despite all the conducted research and the provided efforts in advancing the healthcare sector, there is a strong need to rapidly and efficiently diagnose various diseases. The complexity of some disease mechanisms on one side and the dramatic life-saving potential on the other side raise big challenges for the development of tools for the early detection and diagnosis of diseases. Deep learning (DL), an area of artificial intelligence (AI), can be an informative medical tomography method that can aid in the early diagnosis of gallbladder (GB) disease based on ultrasound images (UI). Many researchers considered the classification of only one disease of the GB. In this work, we successfully managed to apply a deep neural network (DNN)-based classification model to a rich built database in order to detect nine diseases at once and to determine the type of disease using UI. In the first step, we built a balanced database composed of 10,692 UI of the GB organ from 1782 patients. These images were carefully collected from three hospitals over roughly three years and then classified by professionals. In the second step, we preprocessed and enhanced the dataset images in order to achieve the segmentation step. Finally, we applied and then compared four DNN models to analyze and classify these images in order to detect nine GB disease types. All the models produced good results in detecting GB diseases; the best was the MobileNet model, with an accuracy of 98.35%.

Comprehensive AI-assisted tool for ankylosing spondylitis based on multicenter research outperforms human experts

Introduction

The diagnosis and treatment of ankylosing spondylitis (AS) is a difficult task, especially in less developed countries without access to experts. To address this issue, a comprehensive artificial intelligence (AI) tool was created to help diagnose and predict the course of AS.

Methods

In this retrospective study, a dataset of 5389 pelvic radiographs (PXRs) from patients treated at a single medical center between March 2014 and April 2022 was used to create an ensemble deep learning (DL) model for diagnosing AS. The model was then tested on an additional 583 images from three other medical centers, and its performance was evaluated using the area under the receiver operating characteristic curve analysis, accuracy, precision, recall, and F1 scores. Furthermore, clinical prediction models for identifying high-risk patients and triaging patients were developed and validated using clinical data from 356 patients.

Results

The ensemble DL model demonstrated impressive performance in a multicenter external test set, with precision, recall, and area under the receiver operating characteristic curve values of 0.90, 0.89, and 0.96, respectively. This performance surpassed that of human experts, and the model also significantly improved the experts' diagnostic accuracy. Furthermore, the model's diagnosis results based on smartphone-captured images were comparable to those of human experts. Additionally, a clinical prediction model was established that accurately categorizes patients with AS into high-and low-risk groups with distinct clinical trajectories. This provides a strong foundation for individualized care.

Discussion

In this study, an exceptionally comprehensive AI tool was developed for the diagnosis and management of AS in complex clinical scenarios, especially in underdeveloped or rural areas that lack access to experts. This tool is highly beneficial in providing an efficient and effective system of diagnosis and management.

An integrated nomogram combining deep learning, clinical characteristics and ultrasound features for predicting central lymph node metastasis in papillary thyroid cancer: A multicenter study

OBJECTIVE

Central lymph node metastasis (CLNM) is a predictor of poor prognosis for papillary thyroid carcinoma (PTC) patients. The options for surgeon operation or follow-up depend on the state of CLNM while accurate prediction is a challenge for radiologists. The present study aimed to develop and validate an effective preoperative nomogram combining deep learning, clinical characteristics and ultrasound features for predicting CLNM.

MATERIALS AND METHODS

In this study, 3359 PTC patients who had undergone total thyroidectomy or thyroid lobectomy from two medical centers were enrolled. The patients were divided into three datasets for training, internal validation and external validation. We constructed an integrated nomogram combining deep learning, clinical characteristics and ultrasound features using multivariable logistic regression to predict CLNM in PTC patients.

RESULTS

Multivariate analysis indicated that the AI model-predicted value, multiple, position, microcalcification, abutment/perimeter ratio and US-reported LN status were independent risk factors predicting CLNM. The area under the curve (AUC) for the nomogram to predict CLNM was 0.812 (95% CI, 0.794-0.830) in the training cohort, 0.809 (95% CI, 0.780-0.837) in the internal validation cohort and 0.829(95%CI, 0.785-0.872) in the external validation cohort. Based on the analysis of the decision curve, our integrated nomogram was superior to other models in terms of clinical predictive ability.

CONCLUSION

Our proposed thyroid cancer lymph node metastasis nomogram shows favorable predictive value to assist surgeons in making appropriate surgical decisions in PTC treatment.

Evaluation of a computer-aided diagnostic model for corneal diseases by analyzing in vivo confocal microscopy images

Objective

In order to automatically and rapidly recognize the layers of corneal images using in vivo confocal microscopy (IVCM) and classify them into normal and abnormal images, a computer-aided diagnostic model was developed and tested based on deep learning to reduce physicians' workload.

Methods

A total of 19,612 corneal images were retrospectively collected from 423 patients who underwent IVCM between January 2021 and August 2022 from Renmin Hospital of Wuhan University (Wuhan, China) and Zhongnan Hospital of Wuhan University (Wuhan, China). Images were then reviewed and categorized by three corneal specialists before training and testing the models, including the layer recognition model (epithelium, bowman's membrane, stroma, and endothelium) and diagnostic model, to identify the layers of corneal images and distinguish normal images from abnormal images. Totally, 580 database-independent IVCM images were used in a human-machine competition to assess the speed and accuracy of image recognition by 4 ophthalmologists and artificial intelligence (AI). To evaluate the efficacy of the model, 8 trainees were employed to recognize these 580 images both with and without model assistance, and the results of the two evaluations were analyzed to explore the effects of model assistance.

Results

The accuracy of the model reached 0.914, 0.957, 0.967, and 0.950 for the recognition of 4 layers of epithelium, bowman's membrane, stroma, and endothelium in the internal test dataset, respectively, and it was 0.961, 0.932, 0.945, and 0.959 for the recognition of normal/abnormal images at each layer, respectively. In the external test dataset, the accuracy of the recognition of corneal layers was 0.960, 0.965, 0.966, and 0.964, respectively, and the accuracy of normal/abnormal image recognition was 0.983, 0.972, 0.940, and 0.982, respectively. In the human-machine competition, the model achieved an accuracy of 0.929, which was similar to that of specialists and higher than that of senior physicians, and the recognition speed was 237 times faster than that of specialists. With model assistance, the accuracy of trainees increased from 0.712 to 0.886.

Conclusion

A computer-aided diagnostic model was developed for IVCM images based on deep learning, which rapidly recognized the layers of corneal images and classified them as normal and abnormal. This model can increase the efficacy of clinical diagnosis and assist physicians in training and learning for clinical purposes.

[New advances in the diagnosis and treatment of biliary atresia]

The diagnosis of biliary atresia (BA) is mainly based on clinical manifestations, screening, and related biochemistry tests. In recent years, the development of blood biomarkers and the improvement in ultrasound examination have made it possible for BA to be diagnosed at a younger age. In particular, matrix metalloproteinase-7 shows high sensitivity and specificity and has a higher diagnostic efficiency than existing biochemical parameters, thereby holding a promise for clinical application. Sound touch elastography can increase the diagnostic efficiency for BA in terms of diagnosis and prognostic evaluation. Surgery is still the only method for the treatment of BA at present, with the preferred surgical treatment regimen of Kasai portoenterostomy combined with pharmacotherapies for alleviating infection and inflammation, and the patients who fail Kasai portoenterostomy or have liver dysfunction may require liver transplantation to save their lives. Therefore, the current research on BA should focus on the biomarkers for early diagnosis, specifically targeted drugs, and drugs for preventing progressive liver fibrosis. This article reviews the current diagnosis and treatment methods for BA and discusses the potential research directions.

Diagnostic Efficacy of Advanced Ultrasonography Imaging Techniques in Infants with Biliary Atresia (BA): A Systematic Review and Meta-Analysis

The early diagnosis of biliary atresia (BA) in cholestatic infants is critical to the success of the treatment. Intraoperative cholangiography (IOC), an invasive imaging technique, is the current strategy for the diagnosis of BA. Ultrasonography has advanced over recent years and emerging techniques such as shear wave elastography (SWE) have the potential to improve BA diagnosis. This review sought to evaluate the diagnostic efficacy of advanced ultrasonography techniques in the diagnosis of BA. Six databases (CINAHL, Medline, PubMed, Google Scholar, Web of Science (core collection), and Embase) were searched for studies assessing the diagnostic performance of advanced ultrasonography techniques in differentiating BA from non-BA causes of infantile cholestasis. The meta-analysis was performed using Meta-DiSc 1.4 and Comprehensive Meta-analysis v3 software. Quality Assessment of Diagnostic Accuracy Studies tool version 2 (QUADAS-2) assessed the risk of bias. Fifteen studies consisting of 2185 patients (BA = 1105; non-BA = 1080) met the inclusion criteria. SWE was the only advanced ultrasonography technique reported and had a good pooled diagnostic performance (sensitivity = 83%; specificity = 77%; AUC = 0.896). Liver stiffness indicators were significantly higher in BA compared to non-BA patients (p < 0.000). SWE could be a useful tool in differentiating BA from non-BA causes of infantile cholestasis. Future studies to assess the utility of other advanced ultrasonography techniques are recommended.

The diagnostic performance of ultrasound elastography for biliary atresia: A meta-analysis

Background

Biliary atresia (BA) is a severe inflammatory obliterative cholangiopathy of infancy that requires early diagnosis and prompt surgical intervention. In this study, we aimed to obtain comprehensive evidence on the diagnostic performance of liver stiffness measurement by ultrasound elastography in the detection of BA through a meta-analysis.

Methods

The PubMed, EMBASE, Cochrane Library, and Web of Science databases were searched for studies that investigated the diagnostic performance of ultrasound elastography in the detection of BA up to January 10, 2022. In this study, in order to summarize the diagnostic performance of ultrasound elastography, the summary receiver operating characteristic (SROC) modeling was constructed. Heterogeneity was estimated with the I ² statistic. Multiple subgroup analyses were also performed.

Results

Fourteen studies from eleven articles, including 774 BA patients, 850 non-BA patients, and 173 controls were included in the present meta-analysis. The summary sensitivity and specificity of ultrasound elastography for liver stiffness were 85% [95% confidence interval (CI): 79-89%] and 82% (95% CI: 73-88%) with the I ² value of 82.90 and 84.33%, respectively. The area under the SROC curve (AUROC) using ultrasound elastography for diagnosing BA was 0.90 (95% CI: 0.87-0.92). In addition, a subgroup analysis of 9 two-dimensional shear wave elastography studies was also performed. Subgroup analysis revealed that the summary sensitivity and specificity were 85% (95% CI: 77-91%) and 79% (95% CI: 71-86%), respectively, and the summary AUROC was 0.89 (95% CI: 0.86-0.92).

Conclusions

Ultrasound elastography exhibits good diagnostic accuracy for BA and can be served as a non-invasive tool to facilitate the differential diagnosis of BA.

The application of artificial intelligence to support biliary atresia screening by ultrasound images: A study based on deep learning models

Purpose

Early confirmation or ruling out biliary atresia (BA) is essential for infants with delayed onset of jaundice. In the current practice, percutaneous liver biopsy and intraoperative cholangiography (IOC) remain the golden standards for diagnosis. In Taiwan, the diagnostic methods are invasive and can only be performed in selective medical centers. However, referrals from primary physicians and local pediatricians are often delayed because of lacking clinical suspicions. Ultrasounds (US) are common screening tools in local hospitals and clinics, but the pediatric hepatobiliary US particularly requires well-trained imaging personnel. The meaningful comprehension of US is highly dependent on individual experience. For screening BA through human observation on US images, the reported sensitivity and specificity were achieved by pediatric radiologists, pediatric hepatobiliary experts, or pediatric surgeons. Therefore, this research developed a tool based on deep learning models for screening BA to assist pediatric US image reading by general physicians and pediatricians.

Methods

De-identified hepatobiliary US images of 180 patients from Taichung Veterans General Hospital were retrospectively collected under the approval of the Institutional Review Board. Herein, the top network models of ImageNet Large Scale Visual Recognition Competition and other network models commonly used for US image recognition were included for further study to classify US images as BA or non-BA. The performance of different network models was expressed by the confusion matrix and receiver operating characteristic curve. There were two methods proposed to solve disagreement by US image classification of a single patient. The first and second methods were the positive-dominance law and threshold law. During the study, the US images of three successive patients suspected to have BA were classified by the trained models.

Results

Among all included patients contributing US images, 41 patients were diagnosed with BA by surgical intervention and 139 patients were either healthy controls or had non-BA diagnoses. In this study, a total of 1,976 original US images were enrolled. Among them, 417 and 1,559 raw images were from patients with BA and without BA, respectively. Meanwhile, ShuffleNet achieved the highest accuracy of 90.56% using the same training parameters as compared with other network models. The sensitivity and specificity were 67.83% and 96.76%, respectively. In addition, the undesired false-negative prediction was prevented by applying positive-dominance law to interpret different images of a single patient with an acceptable false-positive rate, which was 13.64%. For the three consecutive patients with delayed obstructive jaundice with IOC confirmed diagnoses, ShuffleNet achieved accurate diagnoses in two patients.

Conclusion

The current study provides a screening tool for identifying possible BA by hepatobiliary US images. The method was not designed to replace liver biopsy or IOC, but to decrease human error for interpretations of US. By applying the positive-dominance law to ShuffleNet, the false-negative rate and the specificities were 0 and 86.36%, respectively. The trained deep learning models could aid physicians other than pediatric surgeons, pediatric gastroenterologists, or pediatric radiologists, to prevent misreading pediatric hepatobiliary US images. The current artificial intelligence (AI) tool is helpful for screening BA in the real world.

Using Artificial Intelligence to Analyze Non-Human Drawings: A First Step with Orangutan Productions

Drawings have been widely used as a window to the mind; as such, they can reveal some aspects of the cognitive and emotional worlds of other animals that can produce them. The study of non-human drawings, however, is limited by human perception, which can bias the methodology and interpretation of the results. Artificial intelligence can circumvent this issue by allowing automated, objective selection of features used to analyze drawings. In this study, we use artificial intelligence to investigate seasonal variations in drawings made by Molly, a female orangutan who produced more than 1299 drawings between 2006 and 2011 at the Tama Zoological Park in Japan. We train the VGG19 model to first classify the drawings according to the season in which they are produced. The results show that deep learning is able to identify subtle but significant seasonal variations in Molly's drawings, with a classification accuracy of 41.6%. We use VGG19 to investigate the features that influence this seasonal variation. We analyze separate features, both simple and complex, related to color and patterning, and to drawing content and style. Content and style classification show maximum performance for moderately complex, highly complex, and holistic features, respectively. We also show that both color and patterning drive seasonal variation, with the latter being more important than the former. This study demonstrates how deep learning can be used to objectively analyze non-figurative drawings and calls for applications to non-primate species and scribbles made by human toddlers.

Advanced Image Analysis Methods for Automated Segmentation of Subnuclear Chromatin Domains

The combination of ever-increasing microscopy resolution with cytogenetical tools allows for detailed analyses of nuclear functional partitioning. However, the need for reliable qualitative and quantitative methodologies to detect and interpret chromatin sub-nuclear organization dynamics is crucial to decipher the underlying molecular processes. Having access to properly automated tools for accurate and fast recognition of complex nuclear structures remains an important issue. Cognitive biases associated with human-based curation or decisions for object segmentation tend to introduce variability and noise into image analysis. Here, we report the development of two complementary segmentation methods, one semi-automated (iCRAQ) and one based on deep learning (Nucl.Eye.D), and their evaluation using a collection of A. thaliana nuclei with contrasted or poorly defined chromatin compartmentalization. Both methods allow for fast, robust and sensitive detection as well as for quantification of subtle nucleus features. Based on these developments, we highlight advantages of semi-automated and deep learning-based analyses applied to plant cytogenetics.

Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study

BACKGROUND

Studies on deep learning (DL)-based models in breast ultrasound (US) remain at the early stage due to a lack of large datasets for training and independent test sets for verification. We aimed to develop a DL model for differentiating benign from malignant breast lesions on US using a large multicenter dataset and explore the model's ability to assist the radiologists.

METHODS

A total of 14,043 US images from 5012 women were prospectively collected from 32 hospitals. To develop the DL model, the patients from 30 hospitals were randomly divided into a training cohort (n = 4149) and an internal test cohort (n = 466). The remaining 2 hospitals (n = 397) were used as the external test cohorts (ETC). We compared the model with the prospective Breast Imaging Reporting and Data System assessment and five radiologists. We also explored the model's ability to assist the radiologists using two different methods.

RESULTS

The model demonstrated excellent diagnostic performance with the ETC, with a high area under the receiver operating characteristic curve (AUC, 0.913), sensitivity (88.84%), specificity (83.77%), and accuracy (86.40%). In the comparison set, the AUC was similar to that of the expert (p = 0.5629) and one experienced radiologist (p = 0.2112) and significantly higher than that of three inexperienced radiologists (p < 0.01). After model assistance, the accuracies and specificities of the radiologists were substantially improved without loss in sensitivities.

CONCLUSIONS

The DL model yielded satisfactory predictions in distinguishing benign from malignant breast lesions. The model showed the potential value in improving the diagnosis of breast lesions by radiologists.

A deep learning-based model improves diagnosis of early gastric cancer under narrow band imaging endoscopy

BACKGROUND

Diagnosis of early gastric cancer (EGC) under narrow band imaging endoscopy (NBI) is dependent on expertise and skills. We aimed to elucidate whether artificial intelligence (AI) could diagnose EGC under NBI and evaluate the diagnostic assistance of the AI system.

METHODS

In this retrospective diagnostic study, 21,785 NBI images and 20 videos from five centers were divided into a training dataset (13,151 images, 810 patients), an internal validation dataset (7057 images, 283 patients), four external validation datasets (1577 images, 147 patients), and a video validation dataset (20 videos, 20 patients). All the images were labeled manually and used to train an AI system using You look only once v3 (YOLOv3). Next, the diagnostic performance of the AI system and endoscopists were compared and the diagnostic assistance of the AI system was assessed. The accuracy, sensitivity, specificity, and AUC were primary outcomes.

RESULTS

The AI system diagnosed EGCs on validation datasets with AUCs of 0.888-0.951 and diagnosed all the EGCs (100.0%) in video dataset. The AI system achieved better diagnostic performance (accuracy, 93.2%, 95% CI, 90.0-94.9%) than senior (85.9%, 95% CI, 84.2-87.4%) and junior (79.5%, 95% CI, 77.8-81.0%) endoscopists. The AI system significantly enhanced the performance of endoscopists in senior (89.4%, 95% CI, 87.9-90.7%) and junior (84.9%, 95% CI, 83.4-86.3%) endoscopists.

CONCLUSION

The NBI AI system outperformed the endoscopists and exerted potential assistant impact in EGC identification. Prospective validations are needed to evaluate the clinical reinforce of the system in real clinical practice.

A deep learning model and human-machine fusion for prediction of EBV-associated gastric cancer from histopathology

Epstein–Barr virus-associated gastric cancer (EBVaGC) shows a robust response to immune checkpoint inhibitors. Therefore, a cost-efficient and accessible tool is needed for discriminating EBV status in patients with gastric cancer. Here we introduce a deep convolutional neural network called EBVNet and its fusion with pathologists for predicting EBVaGC from histopathology. The EBVNet yields an averaged area under the receiver operating curve (AUROC) of 0.969 from the internal cross validation, an AUROC of 0.941 on an external dataset from multiple institutes and an AUROC of 0.895 on The Cancer Genome Atlas dataset. The human-machine fusion significantly improves the diagnostic performance of both the EBVNet and the pathologist. This finding suggests that our EBVNet could provide an innovative approach for the identification of EBVaGC and may help effectively select patients with gastric cancer for immunotherapy.

Epstein–Barr virus-associated gastric cancer shows a robust response to immune checkpoint inhibitors. Here the authors introduce a deep convolutional neural network and its fusion with pathologists for predicting it from histopathology.

An Attention-Preserving Network-Based Method for Assisted Segmentation of Osteosarcoma MRI Images

Osteosarcoma is a malignant bone tumor that is extremely dangerous to human health. Not only does it require a large amount of work, it is also a complicated task to outline the lesion area in an image manually, using traditional methods. With the development of computer-aided diagnostic techniques, more and more researchers are focusing on automatic segmentation techniques for osteosarcoma analysis. However, existing methods ignore the size of osteosarcomas, making it difficult to identify and segment smaller tumors. This is very detrimental to the early diagnosis of osteosarcoma. Therefore, this paper proposes a Contextual Axial-Preserving Attention Network (CaPaN)-based MRI image-assisted segmentation method for osteosarcoma detection. Based on the use of Res2Net, a parallel decoder is added to aggregate high-level features which effectively combines the local and global features of osteosarcoma. In addition, channel feature pyramid (CFP) and axial attention (A-RA) mechanisms are used. A lightweight CFP can extract feature mapping and contextual information of different sizes. A-RA uses axial attention to distinguish tumor tissues by mining, which reduces computational costs and thus improves the generalization performance of the model. We conducted experiments using a real dataset provided by the Second Xiangya Affiliated Hospital and the results showed that our proposed method achieves better segmentation results than alternative models. In particular, our method shows significant advantages with respect to small target segmentation. Its precision is about 2% higher than the average values of other models. For the segmentation of small objects, the DSC value of CaPaN is 0.021 higher than that of the commonly used U-Net method.

Medical Professional Enhancement Using Explainable Artificial Intelligence in Fetal Cardiac Ultrasound Screening

Diagnostic support tools based on artificial intelligence (AI) have exhibited high performance in various medical fields. However, their clinical application remains challenging because of the lack of explanatory power in AI decisions (black box problem), making it difficult to build trust with medical professionals. Nevertheless, visualizing the internal representation of deep neural networks will increase explanatory power and improve the confidence of medical professionals in AI decisions. We propose a novel deep learning-based explainable representation "graph chart diagram" to support fetal cardiac ultrasound screening, which has low detection rates of congenital heart diseases due to the difficulty in mastering the technique. Screening performance improves using this representation from 0.966 to 0.975 for experts, 0.829 to 0.890 for fellows, and 0.616 to 0.748 for residents in the arithmetic mean of area under the curve of a receiver operating characteristic curve. This is the first demonstration wherein examiners used deep learning-based explainable representation to improve the performance of fetal cardiac ultrasound screening, highlighting the potential of explainable AI to augment examiner capabilities.

The favorable prognosis of cystic biliary atresia may be related to early surgery and mild liver pathological changes

OBJECTIVE

The objectives of this study is to compare the prognostic differences between cystic biliary atresia (CBA) and non-CBA, analyze the clinical and liver pathological differences between the two groups, and explore the possible factors that affect the native liver survival of infants with CBA after Kasai portoenterostomy (KPE).

METHODS

From 2013 to 2020, 131 infants with BA were admitted to Tianjin Children's Hospital. A total of 108 infants with BA were included after excluding those who did not undergo surgery after diagnosis (n = 23), including 12 cases of CBA and 96 cases of non-CBA. The clinical data, native liver survival and liver pathology, including liver fibrosis, bile ductular proliferation (BDP), bile plug, and portal area inflammation infiltration of the two study groups were compared.

RESULTS

CBA accounts for 9.16% (12/131) and type I CBA accounts for 6.87% (9/131) of all types of BA. 16.7% (2/12) of CBA were detected prenatally with diagnosis of choledochal cyst (CC). The age at KPE, total bilirubin, direct bilirubin, and total bile acid levels of CBA were significantly lower than those of non-CBA (P = 0.047, P = 0.013, P = 0.009, P = 0.010, respectively). The long and wide diameters of the gallbladder were significantly larger than those of non-CBA (both P < 0.001). The 1-, 3-, and 5-year survival rates of CBA were 83.3%, 71.4%, and 71.4%, respectively, and 56.5%, 32.5%, and 29.8%, respectively, in non-CBA. The difference between the two groups was statistically significant (P = 0.031). The degree of liver fibrosis and bile plug in non-CBA was higher than that of CBA (P = 0.004, P < 0.001, respectively). There was no difference of BDP and inflammation infiltration between the two groups (P = 0.285, P = 0.243, respectively).

CONCLUSION

CBA is a distinct type different from non-CBA, with different pathological processes, pathological manifestations, and clinical prognosis. The favorable prognosis of CBA may be derived from the early diagnosis, early operation, and mild pathological changes.

Ultrasound for the Diagnosis of Biliary Atresia: From Conventional Ultrasound to Artificial Intelligence

Biliary atresia is an aggressive liver disease of infancy and can cause death without timely surgical intervention. Early diagnosis of biliary atresia is critical to the recovery of bile drainage and long-term transplant-free survival. Ultrasound is recommended as the initial imaging strategy for the diagnosis of biliary atresia. Numerous ultrasound features have been proved helpful for the diagnosis of biliary atresia. In recent years, with the help of new technologies such as elastography ultrasound, contrast-enhanced ultrasound and artificial intelligence, the diagnostic performance of ultrasound has been significantly improved. In this review, various ultrasound features in the diagnosis of biliary atresia are summarized. A diagnostic decision flow chart for biliary atresia is proposed on the basis of the hybrid technologies, combining conventional ultrasound, elastography and contrast-enhanced ultrasound. In addition, the application of artificial intelligence in the diagnosis of biliary atresia with ultrasound images is also introduced.

Comparison of deep learning systems and cornea specialists in detecting corneal diseases from low-quality images

The performance of deep learning in disease detection from high-quality clinical images is identical to and even greater than that of human doctors. However, in low-quality images, deep learning performs poorly. Whether human doctors also have poor performance in low-quality images is unknown. Here, we compared the performance of deep learning systems with that of cornea specialists in detecting corneal diseases from low-quality slit lamp images. The results showed that the cornea specialists performed better than our previously established deep learning system (PEDLS) trained on only high-quality images. The performance of the system trained on both high- and low-quality images was superior to that of the PEDLS while inferior to that of a senior corneal specialist. This study highlights that cornea specialists perform better in low-quality images than the system trained on high-quality images. Adding low-quality images with sufficient diagnostic certainty to the training set can reduce this performance gap.

•

Deep learning performs poorly in low-quality images for detecting corneal diseases

•

Corneal specialists perform better than the PEDLS in low-quality images

•

The performance of the NDLS is better than that of the PEDLS in low-quality images

•

Adding low-quality images to the training set can improve the system's performance

A systematic review on application of deep learning in digestive system image processing

With the advent of the big data era, the application of artificial intelligence represented by deep learning in medicine has become a hot topic. In gastroenterology, deep learning has accomplished remarkable accomplishments in endoscopy, imageology, and pathology. Artificial intelligence has been applied to benign gastrointestinal tract lesions, early cancer, tumors, inflammatory bowel diseases, livers, pancreas, and other diseases. Computer-aided diagnosis significantly improve diagnostic accuracy and reduce physicians' workload and provide a shred of evidence for clinical diagnosis and treatment. In the near future, artificial intelligence will have high application value in the field of medicine. This paper mainly summarizes the latest research on artificial intelligence in diagnosing and treating digestive system diseases and discussing artificial intelligence's future in digestive system diseases. We sincerely hope that our work can become a stepping stone for gastroenterologists and computer experts in artificial intelligence research and facilitate the application and development of computer-aided image processing technology in gastroenterology.

Machine Learning Evaluation of Biliary Atresia Patients to Predict Long-Term Outcome after the Kasai Procedure

Kasai portoenterostomy (KP) represents the first-line treatment for biliary atresia (BA). The purpose was to compare the accuracy of quantitative parameters extracted from laboratory tests, US imaging, and MR imaging studies using machine learning (ML) algorithms to predict the long-term medical outcome in native liver survivor BA patients after KP. Twenty-four patients were evaluated according to clinical and laboratory data at initial evaluation (median follow-up = 9.7 years) after KP as having ideal (n = 15) or non-ideal (n = 9) medical outcomes. Patients were re-evaluated after an additional 4 years and classified in group 1 (n = 12) as stable and group 2 (n = 12) as non-stable in the disease course. Laboratory and quantitative imaging parameters were merged to test ML algorithms. Total and direct bilirubin (TB and DB), as laboratory parameters, and US stiffness, as an imaging parameter, were the only statistically significant parameters between the groups. The best algorithm in terms of accuracy, sensitivity, specificity, and AUCROC was naive Bayes algorithm, selecting only laboratory parameters (TB and DB). This preliminary ML analysis confirms the fundamental role of TB and DB values in predicting the long-term medical outcome for BA patients after KP, even though their values may be within the normal range. Physicians should be alert when TB and DB values change slightly.

Recent advances in the use of ultrasound and related techniques in diagnosing and predicting outcomes in biliary atresia

PURPOSE OF REVIEW

Biliary atresia (BA) is the leading cause of chronic liver disease and the most common indication for pediatric liver transplantation. The use of ultrasound (US) and related techniques continues to evolve to help diagnose BA as well as potentially to help predict outcomes after treatment with the Kasai portoenterostomy (KP).

RECENT FINDINGS

There are no US findings that are definitive for BA; however, signs which are consistent with BA include gallbladder abnormalities, the triangular cord sign, presence of hepatic subcapsular flow, and hilar lymphadenopathy. Elastography techniques to measure liver stiffness may also increase the diagnostic accuracy of detecting BA, particularly in older infants or without other US findings. In addition, both US and elastography are still being studied as potential methods to predict outcomes after KP such as the development of portal hypertension and the need for liver transplant.

SUMMARY

US findings in the diagnosis of BA are well characterized. Future studies will help determine the utility of elastography in diagnosing BA, as well as both US and elastography in monitoring and predicting disease outcomes after KP.