MOSAIK: A computational framework for theranostic digital twin in renal cell carcinoma

Keywords

Health Analytics
Code:
65/2025
Title:
MOSAIK: A computational framework for theranostic digital twin in renal cell carcinoma
Date:
Monday 27th October 2025
Author(s):
Pottier, A.; Gelardi, F.; Larcher, A.; Capitanio, U.; Rainone, P.; Moresco, R.M.; Tenace, N.; Colecchia, M.; Grassi, S.; Ponzoni, M.; Chiti, A.; Cavinato, L.
Download link:
Abstract:
In nuclear oncology, radiopharmaceuticals (RP) emerged as theranostic tools able to bind specifically to cancer biomarkers and inflict subsequently systematic and irreparable damage to the DNA of the targeted cells. That is why radioisotopes- based therapies entered the clinical practice to diagnose and treat tumours simultaneously and may potentially overcome therapeutic resistance encountered in various cancers. Despite these advancements, tumoural heterogeneity and poor anti- cancer drug penetration in solid tumours turns out to be overlooked pieces of the personalized oncology puzzle, leading to treatment failure. In this study, we propose MOSAIK, an oncological digital twin framework to simulate the intra-tumour uptake of radiopharmaceutical agent, specifically [89Zr]Zr-girentuximab in clear cell Renal Cell Carcinoma (ccRCC). Our comprehensive approach integrates patient-based insights in space and time for reflecting the multi-faceted nature of RP uptake. We develop models to segment blood vessels and identify neoplastic regions, enabling the characterization of the biological domain. To discuss the intra-tumour heterogeneity contribution to the drug diffusion process, we spatially correlate immunochemistry images-derived parameters with the baseline drug accumulation captured through micro PET imaging. Additionally, the model is informed with temporal features leveraged from the compartmental model of the RP agent. The presented Deep Learning (DL) framework incorporates interpretable spatial and temporal inputs stemming from histopathology images. This work aims to provide a computational model with predictive capabilities in drug retention in tissues to move beyond the one-size-fits-all paradigm in nuclear medicine.