Instability as Signal: Diagnosing Latent Heterogeneity in Multimodal Survival Models
- 1MANIPAL UNIVERSITY JAIPUR
Abstract
Multimodal deep learning models for cancer survival prediction frequently exhibit unstable and irreproducible performance, limiting their clinical applicability. Rather than proposing a new predictive architecture, this work presents a diagnostic investigation into the sources of this instability. Using a transformer-based deep fusion model combining RNA sequencing and histopathology data from the TCGA-BRCA cohort as a diagnostic probe, we demonstrate that global survival prediction is highly unstable across data splits. We then analyze the learned multimodal embeddings in an unsupervised manner, revealing stable latent patient subgroups with significantly divergent survival outcomes. These subgroups are not explained by canonical receptor status or immune infiltration, but instead align with tumor-intrinsic proliferative activity. Conditioning survival modeling on the largest discovered subgroup recovers strong and stable prognostic performance, substantially outperforming unimodal baselines. These findings indicate that survival predictability is conditional on latent heterogeneity and suggest that heterogeneity-aware modeling strategies are essential for robust multimodal prognostication. This work establishes a methodological framework for interpreting instability in multimodal survival learning.
Related
- Multimodal spatial-omics reveal the heterogeneity and intercellular network characteristics of papillary craniopharyngiomas.
- Real-world survival outcomes following multimodal surgical treatment for stage IIIA non-small cell lung cancer
- A transcriptional signal associated with mesenchymal stromal cells predicts overall survival in B cell acute lymphoblastic leukemia
- Peritumoral SULF1 Transcript Signal Is Associated with Disease-Specific Survival in Oral Squamous Cell Carcinoma
- Deep learning for multimodal physiological signal based assessment of sleep disordered breathing
BibTeX
@misc{ray2026instability,
title = {{Instability as Signal: Diagnosing Latent Heterogeneity in Multimodal Survival Models}},
author = {R. Ray and S. S. Shekhawat},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.13.26365414},
url = {https://www.medrxiv.org/content/10.64898/2026.10.13.26365414v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Instability as Signal: Diagnosing Latent Heterogeneity in Multimodal Survival Models
AU - R. Ray
AU - S. S. Shekhawat
PY - 2026
DA - 2026/10/13
PB - medRxiv
DO - 10.64898/2026.10.13.26365414
UR - https://www.medrxiv.org/content/10.64898/2026.10.13.26365414v1
AB - Multimodal deep learning models for cancer survival prediction frequently exhibit unstable and irreproducible performance, limiting their clinical applicability. Rather than proposing a new predictive architecture, this work presents a diagnostic investigation into the sources of this instability. Using a transformer-based deep fusion model combining RNA sequencing and histopathology data from the TCGA-BRCA cohort as a diagnostic probe, we demonstrate that global survival prediction is highly unstable across data splits. We then analyze the learned multimodal embeddings in an unsupervised manner, revealing stable latent patient subgroups with significantly divergent survival outcomes. These subgroups are not explained by canonical receptor status or immune infiltration, but instead align with tumor-intrinsic proliferative activity. Conditioning survival modeling on the largest discovered subgroup recovers strong and stable prognostic performance, substantially outperforming unimodal baselines. These findings indicate that survival predictability is conditional on latent heterogeneity and suggest that heterogeneity-aware modeling strategies are essential for robust multimodal prognostication. This work establishes a methodological framework for interpreting instability in multimodal survival learning.
ER -