Back to all papers

GlioSurv: interpretable transformer for multimodal, individualized survival prediction in diffuse glioma.

November 14, 2025pubmed logopapers

Authors

Lee J,Jang J,Eum H,Jang H,Kim M,Park SH,Park CK,Choi SH,Ahn SS,Han Y,Choi KS

Affiliations (12)

  • Interdisciplinary Programs in Cancer Biology, Seoul National University Graduate School, Seoul, Republic of Korea.
  • Department of Biomedical Sciences, Seoul National University, Seoul, Republic of Korea.
  • Department of Radiology, Seoul National University Hospital, Seoul, Republic of Korea.
  • Department of Radiology, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea.
  • Department of Pathology, Seoul National University Hospital, Seoul, Republic of Korea.
  • Department of Neurosurgery, Seoul National University Hospital, Seoul, Republic of Korea.
  • Department of Radiology, Research Institute of Radiological Science, Center for Clinical Imaging Data Science, Yonsei University College of Medicine, Seoul, Republic of Korea.
  • Department of Electronic Engineering, Soongsil University, Seoul, Republic of Korea.
  • Department of Intelligent Semiconductors, Soongsil University, Seoul, Republic of Korea.
  • Department of Radiology, Seoul National University Hospital, Seoul, Republic of Korea. [email protected].
  • Department of Radiology, Seoul National University College of Medicine, Seoul, Republic of Korea. [email protected].
  • Healthcare AI Research Institute, Seoul National University Hospital, Seoul, Korea. [email protected].

Abstract

Adult diffuse gliomas are clinically and molecularly heterogeneous, complicating risk stratification and personalized management. We introduce GlioSurv, a multimodal transformer model based on an accelerated failure time framework to integrate multiparametric MRI, clinical and molecular variables, and treatment data for personalized survival prediction. In a retrospective analysis of 1944 patients, including one internal cohort (n = 891; mean OS 32.2 months) and three external cohorts (n = 84, 470, 499; mean OS 26.1, 18.8, 19.0 months), GlioSurv demonstrated robust discrimination (IAUC: 0.68-0.86), calibration (IBS: 0.10-0.21) and concordance (C-index: 0.61-0.80). It significantly outperformed a convolutional neural network, a vision transformer, and a non-imaging multimodal transformer (p < 0.01). Sequential integration of imaging, clinical, molecular, then treatment data, progressively improved C-index from 0.69 to 0.80 (p < 0.001). Interpretability analyses confirmed established prognostic factors and indicate the potential of GlioSurv to support personalized survival prediction and risk-stratified decision-making in diffuse glioma.

Topics

Journal Article

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.