Prostate cancer is one of the most common types of cancer affecting men, according to World Health Organization (WHO). Current diagnosis and monitoring methods, based on the Prostate-Specific Antigen (PSA) biomarker measured in blood tests exhibit limited precision, often not detecting tumor growth [1]. Advancing the integration of digital twins with physics-based models presents a promising approach to improving personalized diagnosis and supporting clinicians [2]. Here, we propose a machine-learning informed physics-based model combined with a prostate digital twin to reconstruct tumor growth from patient’s PSA blood tests. To achieve that, we generate a three-dimensional prostate representation from patient’s T2-weighted magnetic resonance image (MRI) sequences, incorporating cellular-level details such as cellularity, vascularization, and tumor location. We apply a novel mathematical model based on partial differential equations (PDEs) to simulate PSA levels and tumor growth. The PDE dynamics representing tumor growth are controlled by approximating the spatially and temporarily the fraction of proliferating tumor cells through a deep learning model informed by the digital twin and patient blood tests. We show the potential of the framework with real patient data, demonstrating its capability to reconstruct tumor growth with relative volume errors ranging from 0.8% to 12.28% over a 2.5-year period from diagnosis. Acknowledgements: This publication is part of the project PLEC2021-007709 (ProCanAid), funded by MCIN/AEI/10.13039/501100011033/ and by the European Union NextGenerationEU/PRT and in collaboration with IISLAFE and QUIBIM. It is also supported by the Aragon Regional Government through the grant T50_23R. REFERENCES [1] Holmström, B., Johansson, M., Bergh, A., Stenman, U. H., Hallmans, G., & Stattin, P. (2009). Prostate specific antigen for early detection of prostate cancer: longitudinal study. Bmj, 339. [2] Lorenzo, G., Heiselman, J. S., Liss, M. A., Miga, M. I., Gomez, H., Yankeelov, T. E., ... & Hughes, T. J. (2024). A pilot study on patient-specific computational forecasting of prostate cancer growth during active surveillance using an imaging-informed biomechanistic model. Cancer research communications, 4(3), 617-633.