Multi State Survival Supervised Clustering for Radiomics-Based Cancer Subtyping

This work introduces and evaluate the performance of a novel approach for cancer subtyping using longitudinal data. Our approach leverages a multi state model to stratify patients into subgroups based on their disease progression over time. The method was applied to a longitudinal dataset of patients diagnosed with colorectal liver metastases, aiming to identify distinct subgroups with differing disease-free survival outcomes. The findings suggest that our approach can aid in refining clinical decision-making processes and improving personalized treatment strategies by offering deeper insights into the heterogeneous progression of cancer. Our model shows promise as a tool for advancing cancer subtyping and improving patient management through longitudinal data analysis.