2026-10-07

Purpose: To develop HepaKineticNet, a deep learning framework that models inter-phase kinetic dynamics on multi-phase contrast-enhanced CT (CECT) to generate intratumoral heterogeneity (ITH) maps, and to assess in an independent external cohort whether a clinical-radiomics fusion model improves overall survival (OS) prediction in BCLC stage B-C hepatocellular carcinoma (HCC) treated with locoregional therapy plus first-line TKI and anti-PD-1 agents. Patients And Methods: In this bi-institutional retrospective study, HepaKineticNet was developed using 359 patients from one institution (231 training, 128 validation) and evaluated in an independent external test cohort of 95 patients from a second institution. HepaKineticNet integrates a Spatial Kinetic Differential Attention module into an nnU-Net backbone, separately encoding arterial wash-in and delayed wash-out kinetics, with multiscale coefficient-of-variation self-supervision. Ten radiomics features (hierarchical filtering) and three clinical variables selected by LASSO-penalised Cox regression (10-fold cross-validation) were used to build clinical, radiomics, and fusion Cox models. Results: A total of 277 deaths occurred, 128 of them in the development cohort. Baseline characteristics were balanced (all P>0.05). In the external test cohort, the fusion model achieved a C-index of 0.713, exceeding the clinical model (0.651; P=0.012). Time-dependent AUCs at 1, 2, and 3 years were 0.760 (95%CI 0.651-0.864), 0.838 (95%CI 0.736-0.930), and 0.909 (95%CI 0.686-1.000), whereas those of the clinical model were 0.676, 0.742, and 0.772. Kaplan-Meier analysis showed median OS of 9.5 vs 19.0 months for high- vs low-risk strata (HR 3.02, 95%CI 1.86-4.91; P<0.001). Discrimination was consistent across ten prespecified subgroups. Conclusion: HepaKineticNet-derived ITH maps encode kinetic and spatial information that modestly improves individualized OS prediction and risk stratification in this combination-therapy HCC population, and may provide a noninvasive decision-support adjunct pending prospective validation.

Introduction: Gliomas are the most common primary malignant tumors of the central nervous system, characterized by significant biological heterogeneity and poor prognosis. Multiparametric MRI-based image analysis has shown promise in prognostic assessment; however, effective integration of deep learning (DL), radiomics, and intratumoral heterogeneity (ITH) quantification remains underexplored. This study aims to develop an integrated prognostic model incorporating clinical and multiparametric MRI-derived features and systematically evaluate the predictive performance of different feature combinations. Methods: This retrospective, multicenter study included 183 patients from the UCSF-TCIA cohort as the training set and 109 patients from Chongqing General Hospital as an independent test set. Based on preoperative multiparametric MRI-including T1-weighted imaging (T1WI), T1-weighted contrast-enhanced imaging (T1CE), T2-weighted imaging (T2WI), T2 fluid-attenuated inversion recovery (T2-FLAIR), diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC)-radiomics, DL, and 2D/3D ITH features were extracted. Feature selection was performed using Least Absolute Shrinkage and Selection Operator-Cox (LASSO-Cox) regression, and both single and fusion models were constructed. Feature-level integration was achieved using a two-stage LASSO-Cox regression to leverage the complementary nature of information from different sources fully. Performance was quantified using the Harrell concordance index (C-index) and validated via Kaplan-Meier survival analysis, time-dependent receiver operating characteristic (ROC) analysis, calibration curves, and decision curve analysis (DCA). Results: The fusion model integrating DL, radiomics, ITH, and clinical features (DenseNet121_DL+Rad+ITH) achieved C-indices of 0.801 (95% CI: 0.750-0.853) and 0.710 (95% CI: 0.638-0.783) on the training and test sets, respectively, outperforming all single models. The corresponding stacking ensemble yielded a comparable test C-index of 0.701. Kaplan-Meier analysis confirmed significant risk stratification (P < 0.001). Time-dependent ROC analysis showed AUCs of 0.705, 0.721, 0.746, 0.800, and 0.873 at 12, 18, 24, 30, and 36 months. Calibration curves demonstrated good agreement between predicted and observed survival probabilities. DCA indicated favorable clinical net benefit for medium- to long-term survival prediction (24-36 months). Conclusions: An integrated prognostic model combining clinical characteristics with multiparametric MRI-derived features (DL, radiomics, and ITH) demonstrates strong prognostic potential for glioma patients. Fusion strategies generally outperform single-feature models; however, incremental gains from fusion are modest, and prospective validation in larger cohorts remains necessary.

Background: Heart failure with reduced ejection fraction (HFrEF) links to adverse kidney and cardiovascular events. However, current machine learning (ML)-based algorithms do not include comprehensive echocardiographic features or predict adverse kidney outcomes. Methods: We built ML models to predict 3-year all-cause mortality, major adverse cardiovascular events (MACE), and major adverse kidney events (MAKE) in retrospective HFrEF cohorts. Models were developed from 1,486 patients at Taipei Veterans General Hospital during 2011-2018 and externally validated in 246 patients from the heart-failure-post-acute-care program at Taichung Veterans General Hospital during 2020-2023. Model performance was compared to the traditional Meta-Analysis Global Group in Chronic Heart Failure score (MAGGIC), Logistic Regression (LR), and Kidney-Failure-Risk-Equation (KFRE). Results: For mortality, Ensemble had the highest area under receiver-operating-characteristic curve (AUROC, 0.863; 95% confidence interval [CI], 0.814-0.913; p < 0.001), outperforming MAGGIC (AUROC, 0.741; 95% CI, 0.673-0.810) with net reclassification index (NRI) of 0.294 (95% CI, 0.147-0.440; p = 0.001). For MACE, Ensemble (AUROC, 0.816; 95% CI, 0.769-0.864; p < 0.001) surpassed LR (AUROC, 0.734; 95% CI, 0.677-0.791). For MAKE, Ensemble (AUROC, 0.858; 95% CI, 0.806-0.910; p < 0.001) exceeded KFRE (AUROC, 0.647; 95% CI, 0.560-0.735) with NRI of 0.267 (95% CI, 0.133-0.396; p < 0.001). External validation confirmed good Ensemble discrimination (AUROCs of 0.803, 0.808, and 0.859 for mortality, MACE, and MAKE, respectively). An online tool (https://vghhfrefai.com/) was created for application. Conclusion: Ensemble ML models with detailed clinical and echocardiographic features surpassed conventional risk scores, enabling early detection of major adverse cardiorenal events among HFrEF.

Rationale And Objectives: This study aims to develop and externally validate an interpretable machine-learning model that integrates gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid magnetic resonance imaging (MRI) radiomics, qualitative MRI features, and clinical data to accurately distinguish small (≤2 cm) hepatocellular carcinomas from benign hepatic lesions. Materials And Methods: This retrospective, multicenter cohort included 296 lesions (Center A: 209 [146 training, 63 internal test]; Centers B + C: 87 external validation). Three feature categories were extracted and integrated. A multistep feature selection and model tuning pipeline was performed exclusively within the training set using nested cross-validation to prevent data leakage. Seven machine learning (ML) algorithms were compared. In the external validation set, a reader study was performed to compare the diagnostic performance of the fusion model against that of radiologists. SHapley Additive exPlanations (SHAP) was employed to interpret the optimal model. Results: The fusion model combining three feature categories achieved the best diagnostic performance. In the internal test set, the fusion model yielded an area under the receiver operating characteristic curve (AUC) of 0.913, compared to 0.791 for the radiomics-only model. Its discriminative ability remained robust in an independent external validation cohort (AUC = 0.824), and remained stable after mitigating interscanner variability with ComBat harmonization (AUC = 0.806). The fusion model showed numerically higher performance than junior radiologists and performed comparably to senior radiologists. SHAP analysis identified hepatitis B virus status, alpha-fetoprotein, and specific hepatobiliary phase-derived radiomics features as key predictors. Conclusion: An interpretable ML fusion model demonstrates robust discriminative ability for evaluating small HCCs. Although its numerical advantages over the radiomics-only model and junior radiologists did not reach statistical significance because of the limited external sample size, this noninvasive tool holds significant potential as a reliable diagnostic adjunct with future dynamic threshold optimization.

Aims: To evaluate whether artificial intelligence (AI)-derived retinal vascular parameters provide incremental predictive value for incident type 2 diabetes (T2DM) beyond the American Diabetes Association Risk Test (ADART), and to assess external performance in an independent Chinese community-based cohort. Methods: We analysed the UK Biobank (n = 63,968; 70:30 derivation/internal validation split) and the BRAVE cohort (n = 1159; external validation). Multivariable Cox proportional hazards models with/without AI-derived retinal parameters were constructed to predict the risk of incident T2DM. Model performance was assessed using Harrell's C-index with bootstrap confidence intervals, calibration, continuous net reclassification improvement (NRI) with event and non-event components, integrated discrimination improvement (IDI), and decision curve analysis. Results: Retinal arterial fractal dimension (FDa) and arterial tortuosity (atts) were independently associated with incident T2DM after adjustment for ADART. In the UK Biobank internal validation set, adding FDa and atts increased the C-index from 0.746 to 0.751 (delta = 0.005, 95 % CI 0.003 to 0.008); continuous NRI was 0.258, with event NRI 0.158 and non-event NRI 0.100. In BRAVE, the C-index increased from 0.618 to 0.631 (delta = 0.013, 95 % CI 0.009 to 0.016), supporting external validation of discrimination. Conclusions: AI-derived retinal vascular parameters, particularly FDa and atts, provided measurable incremental predictive information beyond ADART for incident T2DM. These findings support the use of retinal AI biomarkers as complementary risk markers, although prospective validation with large sample sizes, recalibration in target populations, and cost-effectiveness assessment are required before clinical use.

Purpose: To develop and validate an interpretable deep learning pipeline for the automated measurement of the calcaneal inclination angle (CIA) on lateral weight-bearing foot radiographs using partial semantic segmentation. Methods: This retrospective single-center study included 884 radiographs after the removal of duplicates, partitioned into a development cohort of 779 radiographs and a held-out internal test cohort of 105 radiographs, with no patient contributing to both. Three landmark-relevant anatomical regions were manually annotated: the inferior calcaneal contour (CAL), inferior contour of the fifth metatarsal head (M5), and calcaneal aspect of the inferior calcaneocuboid joint (CC). Each region was segmented by its own single-class U-Net model, trained from scratch with an unweighted Dice + Binary Cross-Entropy loss and early stopping on the validation Dice score, and deterministic post-processing was used to derive four anatomical landmarks and to calculate the CIA. A second observer, a fifth-year radiology resident, independently re-annotated a random subset of 50 development radiographs from a blank canvas. The train/validation split was performed at the patient level, and each network was trained with five predefined seeds. Segmentation performance was evaluated on the validation set, and clinical measurement accuracy was assessed on the held-out test set against three readers who measured independently and blinded. The frozen primary model was then applied without modification to an independent external cohort of 50 radiographs from a second institution, measured by the same three readers under the identical blinded protocol. Results: Validation Dice scores across the five seeded training runs were 0.935 ± 0.010 for CAL, 0.871 ± 0.005 for M5, and 0.805 ± 0.009 for CC (mean ± standard deviation), with the pre-specified primary run reaching 0.936, 0.865, and 0.807, respectively. In the held-out test cohort, the model showed a mean absolute error (MAE) of 0.690° (95% confidence interval [CI] 0.525-0.855) against the mean of the three readers, with 92.4% of measurements within 2° and 98.1% within 3°. Bland-Altman analysis demonstrated a small negative bias of -0.425° (95% confidence interval -0.620° to -0.230°), with 95% limits of agreement from -2.427° to +1.577°. The intraclass correlation coefficient (ICC) formula ICC(A,1) achieved 0.986 across the model and the three readers and 0.994 among the three readers alone. In the external cohort, the model achieved an MAE of 0.85° against the three-reader mean, with 92.0% of measurements within 2° and an ICC of 0.969. Conclusion: Partial semantic segmentation combined with deterministic landmark extraction enables rapid and clinically accurate automated CIA measurement on lateral weight-bearing foot radiographs, with an agreement approaching, though not yet equaling, that observed among human readers. Clinical Significance: The proposed pipeline enabled fast, reproducible CIA measurement with high expert agreement and may support standardized foot alignment assessment after prospective validation.