Deep Phenotyping of Prostate Cancer via Multi-Modal Imaging and AI to Discover High-Risk Phenotypes

How to cite: Shiradkar R. “Deep Phenotyping of Prostate Cancer via Multi-Model Imaging and AI to Discover High-Risk Phenotypes.” September 15, 2025. Accessed Oct 2026. https://grandroundsinurology.com/deep-phenotyping-of-prostate-cancer-via-multi-modal-imaging-and-ai-to-discover-high-risk-phenotypes/

Summary

Rakesh Shiradkar, PhD, Indiana University, Indianapolis, Indiana, discusses combining imaging, pathology, omics, and demographic data to improve understanding of tumor progression and identify high-risk phenotypes (deep phenotyping) of prostate cancer through multi-modal imaging and artificial intelligence (AI). Dr. Shiradkar asserts this will enable earlier intervention, personalized therapy, and better prediction of disease trajectories.

Dr. Shiradkar describes prior work showing that AI-derived signatures from T2-weighted imaging and apparent diffusion coefficient (ADC) mapping can identify high-risk disease and improve prognostication of biochemical recurrence. Similar approaches applied to pathology have allowed AI to capture glandular patterns associated with recurrence risk.

Dr. Shiradkar emphasizes emerging foundation models, such as prostate-specific pathology encoders, that can be adapted for prognosis and risk stratification. He notes recent progress in combining multiparametric magnetic resonance imaging (MRI) with pathology and molecular data. He also cites studies integrating MRI with positron emission tomography (PET), which may improve detection when one modality misses lesions visible in the other.

Curating and linking multimodal datasets across institutions remains challenging. Efforts such as intentionally embedding data collection in clinical trials and submitting to public challenges, including those linking whole-mount histology, MRI, and molecular data, are promising strategies. Dr. Shiradkar also discusses the role of generative AI and data imputation in addressing missing modalities, suggesting these techniques allow for more complete multimodal integration.

Dr. Shiradkar shares that multi-modal, multi-scale data integration through AI has the potential to uncover high-risk prostate cancer phenotypes earlier, but notes progress requires coordinated data curation and validation.

The Global Summit on Precision Diagnosis and Treatment of Prostate Cancer is a unique multi-disciplinary forum organized to inform the key health care stakeholders about the emerging advances in clinical case and research and create a consensus-based vision for the future of precision care and educational and research strategy for its realization. The mission of the Summit is to fill the currently existing gap between the key experts of in vivo imaging, the world authorities in the in vitro fluid- and tissue-based molecular diagnostics, including genomics, and thought leaders in the development of novel observation strategies (e.g., active surveillance, or AS) and therapeutic interventions.