4
Background and AimsThe 12-lead electrocardiogram (ECG) remains a cornerstone of cardiac diagnostics, yet existing artificial intelligence (AI) solutions for automated interpretation often lack generalizability, remain closed source, and are primarily trained using supervised learning (SL), which requires extensive labelled datasets and may limit adaptability across diverse clinical settings. Self-supervised learning (SSL) can potentially overcome these limitations by learning robust representations from unlabelled data. To address these challenges, this study developed and compared two open-source foundational ECG models: DeepECG-SL, a supervised multilabel ECG model, and DeepECG-SSL, a self-supervised model.MethodsBoth models were trained on over 1 million ECGs using a standardized preprocessing pipeline and automated free-text extraction from ECG reports to predict 77 cardiac conditions. DeepECG-SSL leveraged unlabelled data through self-supervised contrastive learning and masked lead modelling before fine-tuning for downstream tasks, while DeepECG-SL was trained directly on labelled diagnostic data in an end-to-end fashion. Performance was evaluated across seven private, multilingual healthcare systems and four public ECG repositories, with assessment of fairness by age and sex, and investigation of privacy vulnerabilities as well as memory and compute requirements.ResultsDeepECG-SSL achieved micro-averaged area under the receiver operating characteristic curves (AUROCs) across all 77 cardiac conditions for ECG interpretation of 0.990 [95% confidence interval (CI): 0.990, 0.990] on the internal dataset (MHI-ds), 0.981 (95% CI: 0.981, 0.981) on external public datasets (UKB, CLSA, MIMIC-IV and PTB), and 0.983 (95% CI: 0.983, 0.983) on external private datasets (UW, UCSF, JGH, NYP, MGH, CSH and CHUM), while DeepECG-SL demonstrated AUROCs of 0.992 (95% CI: 0.992, 0.992), 0.980 (95% CI: 0.980, 0.980), and 0.983 (95% CI: 0.983, 0.984), respectively. Fairness analyses revealed minimal disparities (true-positive rate and false-positive rate difference <0.1) across age and sex groups for both models. DeepECG-SSL demonstrated superior performance on limited-data digital biomarker tasks, with the largest improvements in long QT syndrome (LQTS) genotype classification (AUROC 0.931 vs 0.850, P = .026, n = 127 ECGs) and 5 year atrial fibrillation risk prediction (AUROC 0.742 vs 0.734, P < 0.001, n = 132 050 ECGs), while achieving superior performance in left ventricular ejection fraction ≤40% classification (AUROC 0.926 vs 0.917, P < 0.001, n = 25 252 ECGs) and comparable performance in LQTS detection (AUROC 0.767 vs 0.735, P = 0.117, n = 934 ECGs).ConclusionsThis study establishes SSL as a promising paradigm for ECG analysis, particularly in settings with limited annotated data, enhancing accessibility, generalizability, and fairness in AI-driven cardiac diagnostics. By releasing model weights, preprocessing tools, and validation code, this work aims to support robust, data-efficient AI diagnostics across diverse clinical environments and questions.
47Outcome prediction by serum calprotectin in patients with COVID-19 in the emergency department
Outcome prediction by serum calprotectin in patients with COVID-19 in the emergency department
7Artificial intelligence-enhanced six-lead portable electrocardiogram device for detecting left ventricular systolic dysfunction: a prospective single-centre cohort study.
The real-world effectiveness of the artificial intelligence model based on electrocardiogram (AI-ECG) signals from portable devices for detection of left ventricular systolic dysfunction (LVSD) requires further exploration. In this prospective, single-centre study, we assessed the diagnostic performance of AI-ECG for detecting LVSD using a six-lead hand-held portable device (AliveCor KardiaMobile 6L). We retrained the AI-ECG model, previously validated with 12-lead ECG, to interpret the 6-lead ECG inputs. Patients aged 19 years or older underwent six-lead ECG recording during transthoracic echocardiography. The primary outcome was the area under the receiver operating characteristic curve (AUROC) for detecting LVSD, defined as an ejection fraction below 40%. Of the 1716 patients recruited prospectively, 1635 were included for the final analysis (mean age 60.6 years, 50% male), among whom 163 had LVSD on echocardiography. The AI-ECG model based on the six-lead portable device demonstrated an AUROC of 0.924 [95% confidence interval (CI) 0.903-0.944], with 83.4% sensitivity (95% CI 77.8-89.0%) and 88.7% specificity (95% CI 87.1-90.4%). Of the 1079 patients evaluated using the AI-ECG model based on the conventional 12-lead ECG, the AUROC was 0.962 (95% CI 0.947-0.977), with 90.1% sensitivity (95% CI 85.0-95.2%) and 91.1% specificity (95% CI 89.3-92.9%). The AI-ECG model constructed with the six-lead hand-held portable ECG device effectively identifies LVSD, demonstrating comparable accuracy to that of the conventional 12-lead ECG. This highlights the potential of hand-held portable ECG devices leveraged with AI as efficient tools for early LVSD screening.
Read more8Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department.
Identification of patients with acute coronary syndrome requiring coronary revascularization can be challenging due to inconclusive electrocardiogram (ECG) findings or biomarker results. A deep learning model to detect ECG patterns associated with revascularization likelihood was developed, aiming to guide further assessment and reduce diagnostic uncertainty. A convolutional neural network model was trained on 144 691 ED visits from a US cohort (60 ± 19 years; 53% female; 0.6% revascularization), tested on a separate test cohort (n = 35 995), and benchmarked against clinician ECG interpretation and cardiac troponin T (TnT). External validation was performed for the outcomes revascularization and Type 1 myocardial infarction (MI) on 18 673 ED visits from Europe (55 ± 21 years; 49% female; 1.5% revascularization; 1% Type 1 MI). Primary performance metric was area under the receiver operating characteristic curve (AUROC). In the test cohort, the model achieved an AUROC of 0.91 (95% confidence interval [CI] 0.91-0.91), outperforming clinician ECG interpretation (AUROC 0.65, 95% CI 0.54-0.76) and conventional cardiac TnT (AUROC 0.71). In the external validation cohort, ECG model AUROC was 0.81 (95% CI 0.81-0.82) for revascularization, and 0.85 (95% CI 0.84-0.85) for Type 1 MI, compared with 0.70 (95% CI 0.57-0.83) and 0.74 (95% CI 0.56-0.92) for clinician interpretation, and 0.85 and 0.87 for high-sensitivity (hs)-TnT, respectively. The ECG model had higher specificity but lower sensitivity compared with high-sensitivity-troponin T. The model was able to detect revascularization and Type 1 MI with competitive performance, suggesting a potential role to complement current clinical assessment.
Read more4Novel artificial intelligence model using electrocardiogram for detecting acute myocardial infarction needing revascularization.
Rapid myocardial revascularization in patients with acute myocardial infarction (AMI) is essential to improve clinical outcomes. There is still room for improvement in the timely diagnosis of AMI. This study aimed to develop an artificial intelligence (AI) model using electrocardiograms (ECGs) for detecting AMI needing revascularization. A total of 723 389 ECGs from 300 627 patients in the derivation cohort at a single centre between 2013 and 2020, including 5872 patients with AMI (1.95%) who underwent revascularization, were used for model training and internal testing. A transformer-based deep learning model, initially trained on about one million unlabelled ECGs through self-supervised learning, was fine-tuned for AMI detection. The model's final performance was evaluated in the internal test and the external validation set. The external validation was conducted at an independent centre between 2002 and 2020 using 261 429 ECGs from 259 454 patients, including 1095 patients with AMI (0.42%). By integrating self-supervised learning to train the AI model, we enhanced the AMI detection performance, as demonstrated by an increase in the area under the receiver operating characteristic curve (AUROC) from 0.910 (95% CI, 0.904-0.915) to 0.968 (95% CI, 0.965-0.971) in the external validation set. For ST-elevation myocardial infarction and non-ST-elevation myocardial infarction detection, the AUROCs were 0.991 (95% CI, 0.989-0.993) and 0.947 (95% CI, 0.942-0.952) in the external validation set, respectively. This novel ECG-based AI model may be beneficial for the timely identification of patients with AMI needing revascularization.
Read more16Comparison of the AIMS65 score with the Glasgow-Blatchford and Rockall scoring systems for the prediction of the risk of in-hospital death among patients with upper gastrointestinal bleeding.
the aim of this study was to compare the AIMS65, Glasgow-Blatchford score (GBS) and Rockall score for the prediction of the risk of in-hospital death among patients with upper gastrointestinal bleeding (UGIB). patients with UGIB admitted to the ZhongDa hospital from June 2015 to July 2017 were retrospectively collected. All patients were assessed by the AIMS65, GBS and Rockall score and the main outcomes were in-hospital mortality. Odds ratios (OR) and 95 % confidence interval (CI) were estimated to assess the association of the three scores with the risk of death using logistic regression models. Subsequently, their risk stratification accuracy were compared. Finally, their predictive power was compared using the area under the receiver operating characteristic curve (AUROC). of the 284 UGIB patients enrolled in the study, 51 (18.0 %) had variceal bleeding (VUGIB) and 10 patients (3.5 %) died. AIMS65 (OR = 5.14, 95 % CI = 2.48, 10.64), GBS (OR = 1.66, 95 % CI = 1.28, 2.15) and Rockall (OR = 2.72, 95 % CI = 1.76, 4.18) scores were positively associated with death risk among all patients. The AIMS65 score (high-risk group vs low-risk group: 11.9 % vs 0.0 %, p < 0.001) was effective to classify high-risk in-hospital deaths populations. The AIMS65 score was the best approach to predict in-hospital death among all UGIB patients (AUROC: AIMS65 0.955, GBS 0.882, Rockall 0.938), NVUGIB patients (AUROC = 0.969, 95 % CI = 0.937, 0.989) or VUGIB patients (AUROC = 0.885, 95 % CI = 0.765, 0.967). the AIMS65 score is the most convenient UGIB prognostic score to predict in-hospital mortality and may be more suitable for patients with NVUGIB.
Read more3Ratio of visceral fat area to body fat mass (VBR) is a superior predictor of coronary heart disease.
Ratio of visceral fat area to body fat mass (VBR) is a superior predictor of coronary heart disease.
1Baseline ¹8F-FDG PET/CT radiomics predict early progression and survival in small cell lung cancer.
To determine whether baseline ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) radiomic markers of thoracic tumor burden and metabolic heterogeneity predict early progression and overall survival (OS) in patients with small cell lung cancer (SCLC) receiving first-line platinum-etoposide chemotherapy. We retrospectively analyzed 45 patients with SCLC who underwent baseline ¹⁸F-FDG PET/CT before chemotherapy. Radiomic features were extracted from thoracic tumor volumes, including metabolic tumor volume (MTV), total lesion glycolysis (TLG), standardized uptake value (SUV) histogram parameters, and gray-level co-occurrence matrix (GLCM) metrics. Univariable and multivariable Cox regression assessed associations with progression-free survival (PFS) and OS. Receiver operating characteristic (ROC) analysis, with area under the ROC curve (AUC), assessed discriminatory ability for early progression (PFS < 6 months) and short-term mortality (OS < 12 months). Higher log-transformed TLG (logTLG) was correlated with shorter PFS and OS in univariable analysis. In multivariable models, logTLG independently predicted PFS (hazard ratio [HR] 2.20, 95% confidence interval [CI] 1.12-4.30; p = 0.021) and OS (HR 2.54, 95% CI 1.24-5.20; p = 0.011). Age (HR 1.07 per year; 95% CI 1.02-1.12, p = 0.007) and stage IV vs. III (HR 2.46, 95% CI 1.06-5.71; p = 0.037) were additional OS predictors. GLCM texture features were not significant. ROC analysis showed MTV (AUC 0.88) and serum lactate dehydrogenase (LDH) (AUC 0.74) best predicted early progression, while age (AUC 0.79), stage (AUC 0.73), and SUV entropy (AUC 0.75) predicted short-term mortality. Kaplan-Meier curves confirmed poorer survival with high logTLG, advanced stage, and older age. Baseline thoracic PET/CT radiomics, particularly logTLG, independently predict survival in SCLC. MTV and LDH identify early progression, while age, stage, and SUV entropy predict short-term mortality, supporting integrated imaging-clinical risk stratification.
Read morePredictive scores for critically ill obstetric patients in a resource‐limited setting: A retrospective validation of the Obstetric Early Warning Score.
Background. In South Africa (SA), an unacceptably high institutional maternal mortality rate persists due to failure to recognise critically ill patients. Early warning systems could assist in identifying these patients sooner. Objectives. We evaluated the Obstetric Early Warning Score (OEWS) as a predictor of maternal outcomes in an intensive care unit (ICU) and compared its prognostic validity with the Acute Physiology and Chronic Health Evaluation (APACHE) II and the quick Sequential Organ Failure Assessment score (qSOFA). Methods. Data were extracted on pregnant and post-partum ICU-admitted patients at a tertiary and regional centre in SA between October 2015 and April 2020. Clinical characteristics and outcomes were used to compare the three scoring systems. Results. Among 251 eligible patients, the mortality rate was 8.5%. The OEWS score failed to differentiate between survivors and non-survivors (odds ratio 1.13, 95% confidence intervals (CI) 0.972 - 1.311, p=0.113). APACHE II outperformed the OEWS (area under the receiver operating characteristics curve (AUROC) 0.69, 95% CI 0.540 - 0.846 v. 0.55, 95% CI 0.430 - 0.674). The OEWS (AUROC 0.55, 95% CI 0.430 - 0.674) and qSOFA (AUROC 0.60, 95% CI 0.500 - 0.703) showed no differences. Further analysis revealed that positive scoring for diastolic blood pressure and high systolic blood pressure weakened OEWS performance. Removing these variables improved OEWS prediction (AUROC 0.68). Conclusion. In a SA obstetric population, OEWS did not predict mortality in ICU-admitted patients and offered no advantages over APACHE II or qSOFA scores. Further research should identify critical outcome predictors for low-resource populations.
Read moreIntegrating electrocardiogram into the Mayo score enhances genotype prediction in hypertrophic cardiomyopathy
Introduction Hypertrophic cardiomyopathy (HCM) is a common inherited cardiovascular disease, typically transmitted in an autosomal dominant manner. Although pathogenic variants in sarcomere-encoding genes are key determinants of disease onset and progression, genetic testing has faced hurdles in terms of human and financial resources, necessitating efficient prioritization. The Mayo HCM Genotype Predictor Score (Mayo score) and the Toronto HCM Genotype Score (Toronto score), both derived from clinical and echocardiographic variables, have showed acceptable performance estimating the pre-test probability of genotype positivity. While electrocardiogram (ECG) is a low-cost and minimally invasive modality and pathogenic variants are associated with electrophysiological abnormalities, ECG parameters were not incorporated into these models. Purpose This study aimed to develop an improved predictive model for genotype positivity in HCM by integrating ECG parameters. Methods We retrospectively analysed 507 unrelated HCM patients from a Japanese multicentre cohort. Genotype positivity was defined as the presence of pathogenic or likely pathogenic (P/LP) variants in definitive sarcomeric genes. Candidate ECG variables were selected via univariate analysis and stepwise selection based on Akaike’s information criterion in a multivariable logistic regression model. A novel point-based Mayo-ECG score was developed and internally validated using bootstrap resampling. Predictive performance was assessed by the area under the receiver operating characteristic curve (AUROC) and net reclassification improvement (NRI). Results Overall, 162 (32.0%) were genotype positive. P/LP variants were most frequently identified in MYBPC3 (n = 74), followed by MYH7 (n = 62), and the other genes. Genotype-positive patients had a higher prevalence of atrial fibrillation, intraventricular conduction disturbance (QRS duration ≥ 120 msec or bundle branch block), prolonged QT interval, left axis deviation, lower Sokolow-Lyon index and less frequent T-wave inversion in the precordial leads compared to genotype-negative patients. The Mayo-ECG score (Figure 1) stratified genotype positivity from 7.6% (score ≤ -1) to 90.9% (score ≥ 4) (Figure 2A). Bootstrap resampling demonstrated the Mayo-ECG score (AUROC, 0.81 [95% CI, 0.77-0.85]) outperformed the Mayo score (AUROC, 0.76 [95% CI, 0.72-0.81]; p &lt; 0.001, NRI: 0.21 [95% CI, 0.13-0.29]; p = 0.007) and the Toronto score (AUROC, 0.75 [95% CI, 0.70-0.79]; p = 0.009, NRI: 0.17 [95% CI, 0.06-0.27]; p &lt; 0.001) (Figure 2B), with good calibration (Brier score: 0.156, max calibration error: 0.024). Conclusions Genotype-positive and genotype-negative HCM patients exhibited distinct ECG characteristics. The Mayo-ECG score significantly enhances genotype prediction in HCM by incorporating ECG parameters, outperforming conventional models. Its simplicity and improved accuracy support its use for optimizing genetic screening in HCM.Novel Mayo-ECG score Performance of Mayo-ECG score
Read more50Comprehensive evaluation of 68Ga-PSMA-11 PET/CT parameters for discriminating pathological characteristics in primary clear-cell renal cell carcinoma.
To evaluate parameters derived from 68Ga-PSMA-11 PET/CT images for discriminating pathological characteristics in primary clear-cell renal cell carcinoma (ccRCC). The study retrospectively examined data for 36 ccRCC patients with preoperative 68Ga-PSMA-11 PET/CT scan and surgical specimens. Radiological parameters including maximal tumor diameter, mean CT value, and maximal standard uptake value (SUVmax) were derived from PET/CT images. Pathological characteristics included WHO/ISUP grade and adverse pathology (tumor necrosis or sarcomatoid or rhabdoid feature). Values of radiological parameters were compared within subgroups of pathological characteristics. Receiver operating characteristic (ROC) curve analysis was used for the effectiveness of radiological parameters in differentiating pathological characteristics, estimating area under the ROC curve (AUC) and 95% confidence intervals (CIs). The WHO/ISUP grade distribution for 36 tumors was grade 1, 9 (25.0%); grade 2, 12 (33.3%); grade 3, 9 (25.0%); and grade 4, 6 (16.7%). Adverse pathology was positive for 15 (41.7%). Radiological tumor diameter and SUVmax significantly differed by WHO/ISUP grade, pT stage, and adverse pathology (all P < 0.05), with no difference by CT value. Tumor diameter demonstrated sensitivity 86% and specificity 88% for pT stage, with cutoff 6.70 and AUC 0.91 (95% CI, 0.79-1.00, P< 0.001). SUVmax could effectively differentiate WHO/ISUP grade (3-4 vs. 1-2) and adverse pathology (positive vs. negative), with AUC 0.89 (95% CI, 0.81-0.98, P< 0.001), cutoff 16.4, sensitivity 100%, and specificity 71% and AUC 0.92 (95% CI, 0.85-0.99, P< 0.001), cutoff 18.5, sensitivity 94%, and specificity 87%, respectively. 68Ga-PSMA-11 PET/CT could effectively identify aggressive pathological features of ccRCC.
Read more7Phosphatidylethanol and ethyl glucuronide to categorize alcohol consumption in alcohol-related cirrhosis.
Phosphatidylethanol and ethyl glucuronide to categorize alcohol consumption in alcohol-related cirrhosis.
Comparison of Artificial Intelligence and Radiologists in MRI-Based Prostate Cancer Diagnosis: A Meta-Analysis of Accuracy and Effectiveness
Background: Prostate cancer remains a leading cause of mortality in men, making early, accurate detection crucial for early intervention. While radiologists utilize the Prostate Imaging Reporting and Data System (PI-RADS) for the interpretation of MRI imaging, variations in expertise and inter-reader differences can affect diagnostic accuracy. Artificial intelligence (AI) has emerged as a promising tool for automated detection, with the potential to achieve diagnostic performance comparable to radiologists in identifying clinically significant prostate cancer (csPCa), streamline workflows, and reduce unnecessary biopsies. However, its real-world performance compared to expert radiologists remains a topic of ongoing debate. Purpose: This meta-analysis aims to evaluate whether AI can achieve diagnostic performance that is comparable to that of radiologists in MRI-based prostate cancer detection by comparing diagnostic accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC). Methods: Following PRISMA 2020 guidelines, we searched PubMed for studies directly comparing AI and radiologists in MRI-based detection of csPCa. Ten studies (20,423 patients) were included, and quality was assessed using QUADAS-2. Analyses included forest plots for diagnostic sensitivity and specificity, funnel plots of AUROC to assess publication bias, and paired AUROC difference plots to directly compare diagnostic accuracy. Results: Pooled sensitivity was 0.87 (95% CI: 0.81–0.94) for AI and 0.85 (95% CI: 0.77–0.94) for radiologists; pooled specificity was 0.61 (95% CI: 0.51–0.72) for AI and 0.63 (95% CI: 0.54–0.71) for radiologists. Funnel plots of AUROC against standard error showed no strong visual evidence of publication bias. Paired AUROC difference analysis demonstrated no significant performance difference between AI and radiologists, with a pooled difference of 0.018 (p = 0.378). Conclusions: AI systems demonstrated diagnostic performance comparable to radiologists for MRI-based detection of csPCa, with a nonsignificant and slightly higher pooled sensitivity and AUROC. Moreover, AI has the potential to improve workflow speed, uniformity across expertise levels, and hybrid AI-radiologist approaches to reduce unnecessary biopsies. Large-scale, prospective trials with standardized protocols are needed to assess AI’s effectiveness across diverse clinical settings.
Read more3Mechanical power normalisation methods to predict ICU mortality: a retrospective cohort study
BackgroundThe optimal mechanical ventilation strategy to minimise ventilator-induced lung injury (VILI) remains uncertain. Mechanical power (MP) is a key VILI determinant, but whether and how MP should be normalised to individual patient characteristics is unclear. In this study, we aimed to evaluate whether the discriminatory accuracy of MP for ICU mortality in mechanically ventilated patients improves when normalised to physiologically relevant variables that reflect individual susceptibility to VILI. We also explored whether the relationship between MP, MPratio, and mortality is linear or exhibits a threshold effect.MethodsIn this retrospective observational study, we extracted granular electronic healthcare record data for mechanically ventilated adults in a single centre over a seven-year period. Primary exposures were MP with five normalisations: for dead space (expressed as corrected minute ventilation, ventilatory ratio, or end-tidal to arterial CO2 ratio); aerated lung size (compliance), and normal idealised MP (MPratio). We used logistic regression to assess associations with ICU mortality. We calculated the Area Under the Receiver Operating Characteristic Curve (AUROC) to compare discriminative accuracy of individual models. Additionally, we evaluated the linearity or presence of a threshold for the relationships between MP, MPratio and ICU mortality.ResultWe included 3,578 patients in our analyses. We found MP normalised to compliance (AUROC 0.71, 95% confidence interval (CI) 0.69–0.73, p = 0.007 (DeLong’s test)) and MPratio (AUROC 0.71, 95% CI 0.68–0.73, p = 0.0014) performed better than MP alone (AUROC 0.69, 95% CI 0.66–0.71) for predicting ICU mortality. Other methods of MP normalisation were no more discriminative than MP without normalisation. The relationship between MP and MPratio with ICU mortality showed a statistically significant but small departure from linearity.ConclusionsMechanical power normalised to compliance and MPratio had better discrimination for ICU mortality than MP, although the difference was modest and absolute predictive power remained limited.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13613-025-01562-9.
Read more22The comparison of prognostic value of tumour volumetric regression ratio and RECIST 1.1 criteria after induction chemotherapy in locoregionally advanced nasopharyngeal carcinoma.
The comparison of prognostic value of tumour volumetric regression ratio and RECIST 1.1 criteria after induction chemotherapy in locoregionally advanced nasopharyngeal carcinoma.
Read more7Application of machine learning and deep learning in metabolic dysfunction-associated steatotic liver disease: a systematic review and meta-analysis.
Metabolic dysfunction-associated steatotic liver disease (MASLD) can progress to metabolic dysfunction-associated steatohepatitis (MASH) and liver fibrosis, contributing to a heavier global health burden. Non-invasive diagnostic tools developed using machine learning (ML) and deep learning (DL), two representative artificial intelligence algorithms, are increasingly being explored for MASH and its related fibrosis assessment. This study aimed to compare the diagnostic performance of different ML and DL models and identify the top-performing models for diagnosing MASH and associated liver fibrosis. A systematic review and meta-analysis were conducted across PubMed, Web of Science, Embase and Cochrane Library from inception to May 18, 2025. Pooled area under the receiver operator characteristic curve (AUROC) values with 95 % confidence interval (CI) were calculated. Accuracy, specificity, sensitivity, positive predictive values, and negative predictive values were also recorded. Of 4,314 studies initially identified, 106 met the inclusion criteria, with 35 studies (ML: n=28; DL: n=7) providing data for analysis. Logistic Regression and Neural Network are the most commonly algorithms applied in ML and DL, respectively. The pooled AUROCs for diagnosing MASH were 0.833 (95 %CI: 0.806-0.860) for ML models and 0.841 (95 %CI: 0.782-0.900) for DL models. Light Gradient Boosting Machine (LightGBM) and ResNet50 were the best-performing models for diagnosing MASH within ML and DL algorithms, respectively, achieving corresponding AUROCs of 0.920 (95 %CI: 0.916-0.924) and 0.960 (95 %CI: 0.951-0.969). For fibrosis diagnosis, ML models had a pooled AUROC of 0.826 (95 %CI: 0.792-0.860), with Categorical Boosting (CatBoost) achieving the highest AUROC of 0.960 (95 %CI: 0.950-0.970). DL models yielded the pooled AUROC of 0.875 (95 %CI: 0.816-0.934) for fibrosis diagnosis. Both ML and DL models demonstrated strong diagnostic performance for MASH and liver fibrosis, with DL achieving marginally higher AUROCs. AI-driven approaches show promise in MASLD management.
Read more
Copyright 2026 Cactus Communications. All rights reserved.
