Fusing imaging and metabolic modeling via multimodal deep learning in ovarian cancer

We generate patient-specific metabolic models from transcriptomics data and fuse them with 3D CT imaging using multimodal deep learning. Combining transcriptomics, fluxomics and imaging improves survival estimation, correctly stratifies patients by risk, and gives a mechanistic interpretation of the predictions. The approach can be applied to other cancers with paired imaging and transcriptomics data.

Code: GitHub

Recommended citation: N. Eftekhari, S. Verma, A. Saha, G. Zampieri, S. Sawan, A. Occhipinti, C. Angione, "Fusing imaging and metabolic modeling via multimodal deep learning in ovarian cancer," Cell Systems, vol. 17, 101594, 2026.
Download Paper