End-to-End Opportunistic CT Body Composition Profiling & Interpretable Diabetes Risk Prediction Suites

The software has a built-in Chinese healthy population reference distribution (Kong et al., 2022), supports the generation of one page A4 academic three line table reports (PDF), and has secondary capture (SC) encapsulation that complies with DICOM 3.0 standards and a one click PACS return function based on C-STORE services. App 2: CT body composition risk prediction of diabetes and SHAP interpretable Kanban: combined with multi center retrospective cohort and multi factor logistic regression model, the risk probability of type 2 diabetes in patients was automatically evaluated according to the input body composition characteristics.

The Kanban board integrates individualized dynamic column plots (Waterfall Plot) based on SHAP (Shapley Additive exPlanations) and global feature influence distribution plots (Beeswarm Plot), achieving complete transparency attribution of metabolic phenotype push-pull effects. Pre trained model weights and benchmark dataset: including optimized Total Segmentor L3 localization weights, 2D nnU Net four tissue segmentation checkpoints (checkpoint_final.pth, dataset.rson, plans. json), and a benchmark data table (shap. csv) for generating global SHAP feature distributions. This software suite supports single machine portable operation and edge deployment of medical institutions' local area networks (LANs), and is compatible with Windows 10/11 64 bit environments.

The calculation indicators and risk assessment of this resource are based on specific scientific research queues and are only for reference in medical imaging informatics research, algorithm technology reproduction, and opportunistic screening evaluation. They do not constitute direct clinical medical diagnostic basis.