A Flexible Deep Learning Framework for Survival Analysis with Medical Data

Gabriele Campanella, Ida Häggström, Lucas Kook, Torsten Hothorn, Thomas J. Fuchs

Abstract: Medical imaging data and electronic health records are an integral part of clinical routine and research for prognostication of patient survival and thus directly inform patient management. However, standard regression models used to derive patient prognoses are ill-equipped to handle such non-tabular data directly. Several neural network architectures based on classification or the Cox model have been proposed. Here, we present deep conditional transformation models (DCTMs) for survival applications with medical imaging data. DCTMs include the Cox model as a special case, but parameterize the log cumulative baseline hazards via Bernstein polynomials and allow the specification of non-linear and non-proportional hazards for both tabular and non-tabular data and extend to all types of uninformative censoring. DCTMs yield moderate to large performance gains over state-of-the-art deep learning approaches to survival analysis on a multitude of publicly available datasets featuring tabular or imaging data from radiology and pathology.

External IDs:dblp:conf/miccai/CampanellaHKHF25