A Two-Phase Contrastive Learning Framework for Multi-Modal Ovarian Cancer Prognosis with Explainable AI

DOI

10.2991/978-94-6239-693-7_29How to use a DOI?

Keywords

Ovarian cancer; survival analysis; recurrence prediction; multi-omics; supervised contrastive learning; multi-modal fusion; contrastive learning

Abstract

Ovarian cancer remains one of the most lethal gynecologic malignancies due to late diagnosis and frequent recurrence after treatment. Accurate prediction of both survival and recurrence risk is essential for personalised clinical decision-making. This research work proposes a Multi-Modal Supervised Contrastive Learning framework (MMSCL) that integrates clinical variables, multi-omics data, and CT imaging from 92 patients in the TCGA-OV cohort. The model is trained in two phases. First, supervised contrastive pre-training structures the embedding space using a dual-label pairing strategy based on survival and recurrence outcomes. Second, multi-task fine-tuning simultaneously predicts survival status, recurrence risk, and time-to-recurrence. Dimensionality reduction using PCA and mutual-information-based feature selection reduces noise in the small cohort. Under five-fold cross-validation, MMSCL achieves a mean combined AUC of 0.794±0.040. On an independent hold-out set, the framework obtains a survival AUC of 0.8750 and a Harrell’s C-Index of 0.7091. SHAP analysis highlights tumour stage and molecular subtype features as the most influential predictors. The proposed framework demonstrates the potential of multimodal contrastive learning for interpretable cancer prognosis in small clinical cohorts.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - C. R. Lakshmi
AU  - D. V. Venkata Vara Prasad
AU  - Y. V. Lokeswari
AU  - V. Ramesh
PY  - 2026
DA  - 2026/06/16
TI  - A Two-Phase Contrastive Learning Framework for Multi-Modal Ovarian Cancer Prognosis with Explainable AI
BT  - Proceedings of the International Conference on Intelligent Systems for a Sustainable Future (ISSF 2026)
PB  - Atlantis Press
SP  - 276
EP  - 291
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6239-693-7_29
DO  - 10.2991/978-94-6239-693-7_29
ID  - Lakshmi2026
ER  -