Background: Clear Cell Renal Cell Carcinoma (ccRCC), the most prevalent and aggressive histopathological subtype of kidney cancer, originates from the cells lining the proximal tubules of the nephron. ccRCC tumors show significant intra-tumoral heterogeneity, which impairs drug penetration and therapeutic efficacy. Radiopharmaceutical (RP) therapies have emerged as a promising approach for treating those resistant cancers as they bind specifically to molecular biomarkers expressed in malignant cells.

Methods: In this regard, to move beyond the one-size-fits-all paradigm in nuclear oncology, we developed an oncological digital twin for predicting the intra-tumor uptake of RPT, specifically [89Zr] Zr-girentuximab in RCC. The proposed Deep Learning model captures various mechanisms underlying tumor response from histological sections and the compartmental model of the RP agent. Global temporal dynamics of drug penetration were inferred from the immuno-kinetic compartmental model of [89Zr] Zr-girentuximab. Spatial drug distribution was resolved via tissue characterization, including segmentation of blood vessels and neoplastic regions. Additionally, the model incorporates analyses of proxies for tumoral heterogeneity: immunohistochemistry-derived parameters (e.g., biomarker expression). Spatial correlation techniques were used to identify parameters unraveling the drug uptake patterns in space.

Results: Our study primarily identified key parameters linking tumoral heterogeneity to uneven drug distribution. Thereafter, the multimodal tumor digital twin revealed high-fidelity predictive capabilities in RP drug retention validated against ex-vivo microPET imaging.

Conclusion: The framework adopts a comprehensive approach to account for various aspects of RP therapy absorption in tumors including macroscopic heterogeneity measurements. This patient-specific Digital Twin paves the way for the predictive comparison of treatment efficacy enabling therapeutic optimization and improved clinical outcomes.