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

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Not quantified in abstract — described qualitatively as 'significantly improves model accuracy' compared with transcriptomics-imaging approaches

• Primary Outcome: Survival estimation and binary risk stratification (high vs. low risk) in ovarian cancer patients using multimodal deep learning integrating metabolic models and CT imaging

• Nnt Or Sensitivity: No C-statistic, hazard ratio, sensitivity, specificity, or concordance index reported in the abstract; quantitative performance metrics are unavailable for appraisal at this level

• Confidence Interval: Not reported

Clinical Application

Currently low for routine clinical implementation. The pipeline requires: (1) tumour RNA sequencing for transcriptomics, (2) 3D CT imaging with standardised acquisition, and (3) computational genome-scale metabolic modelling — none of which are co-integrated in standard oncology workflows. Translation would require prospective data collection infrastructure, computational resource investment, and clinical validation trials. The approach is best positioned as a research tool for biomarker discovery and hypothesis generation in the near term. In Australia, ovarian cancer management follows Cancer Australia and RANZCOG guidelines, with PARP inhibitors (olaparib, niraparib) listed on the PBS for BRCA-mutated and HRD-positive patients. The TGA has not evaluated any metabolic-imaging deep learning prognostic tool for ovarian cancer. RACGP and COSA guidelines emphasise FIGO staging and CA-125 for prognostication. This tool, if validated, could complement BRCA/HRD testing by identifying metabolic vulnerabilities and refining risk stratification beyond current biomarkers. Australian implementation would require NHMRC-funded prospective validation using datasets from institutions such as the Peter MacCallum Cancer Centre or the Australian Ovarian Cancer Study (AOCS) biobank, which holds coupled molecular and clinical data. Medicare reimbursement for tumour transcriptomics in ovarian cancer is not currently established, representing a significant access barrier. Patients with ovarian cancer (predominantly high-grade serous ovarian carcinoma, given typical dataset availability) for whom both CT imaging and tumour transcriptomics data are obtainable — currently a research rather than routine clinical population

Abstract

Integrating genotype (e.g., transcriptomics), phenotype (e.g., imaging), and tumor microenvironment (e.g., metabolomics) is crucial to elucidating the molecular basis of ovarian cancer. However, there is a lack of robust multimodal integration methods when only a limited number of common samples is available. Here, we generate patient-specific metabolic models starting from transcriptomics data and integrate them with imaging data. We show that this multimodal integration-never attempted before-improves survival estimation and enables a mechanistic interpretation of the predictions. We assess the robustness of our approach with different combinations of transcriptomics, fluxomics, and 3D computerized tomography (CT) imaging data, correctly stratifying patients based on risk. Fusing metabolic modeling with imaging and transcriptomics significantly improves model accuracy compared with widely used transcriptomics-imaging approaches and elucidates critical metabolic reactions. Our approach is general and can be applied to other cancer types where coupled imaging-transcriptomics data are available. A record of this paper's transparent peer review process is included in the supplemental information.

References

1. 1.Eftekhari, N., Verma, S., Saha, A., Zampieri, G., Sawan, S., Occhipinti, A., & Angione, C. (2026). Fusing imaging and metabolic modeling via multimodal deep learning in ovarian cancer. Cell Systems, 101594. https://doi.org/10.1016/j.cels.2026.101594