Saved in:
Bibliographic Details
Main Authors: Shi, Qiyuan, Li, Yi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.21056
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908987939618816
author Shi, Qiyuan
Li, Yi
author_facet Shi, Qiyuan
Li, Yi
contents Vision Transformers (ViTs) have shown strong empirical performance on high-dimensional medical imaging data, yet their behavior under survival objectives and the interpretability of their attention mechanisms remain poorly understood. Under shallow ViTs, we design controlled experiments showing that token-level attention dynamics can recover outcome-relevant regions and that attention-based thresholding enables effective token pruning, improving both interpretability and predictive performance. We also study pretrained deep ViTs for survival analysis and propose a radiomics-guided hybrid model that integrates pixel-based embeddings with interpretable radiomic features through a multimodal Cox framework and contrastive alignment. Applied to a COVID-19 chest X-ray cohort with a composite ICU admission or mortality endpoint, the proposed approach achieves competitive discrimination while providing clinically meaningful attention maps and feature-group importance.
format Preprint
id arxiv_https___arxiv_org_abs_2604_21056
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Radiomics-Guided Vision Transformers for Survival Analysis
Shi, Qiyuan
Li, Yi
Medical Physics
Vision Transformers (ViTs) have shown strong empirical performance on high-dimensional medical imaging data, yet their behavior under survival objectives and the interpretability of their attention mechanisms remain poorly understood. Under shallow ViTs, we design controlled experiments showing that token-level attention dynamics can recover outcome-relevant regions and that attention-based thresholding enables effective token pruning, improving both interpretability and predictive performance. We also study pretrained deep ViTs for survival analysis and propose a radiomics-guided hybrid model that integrates pixel-based embeddings with interpretable radiomic features through a multimodal Cox framework and contrastive alignment. Applied to a COVID-19 chest X-ray cohort with a composite ICU admission or mortality endpoint, the proposed approach achieves competitive discrimination while providing clinically meaningful attention maps and feature-group importance.
title Radiomics-Guided Vision Transformers for Survival Analysis
topic Medical Physics
url https://arxiv.org/abs/2604.21056