Physics-informed machine learning digital twin for reconstructing prostate cancer tumor growth via PSA tests - Article summary

Aug 5, 2026

Upload by Quinn

Prostate cancer monitoring based on PSA blood tests can miss tumor growth. A computational framework is presented to reconstruct tumor growth using physics-based modeling and machine learning within a digital-twin setting. The physics model links PSA secretion and tissue-to-blood flux to local vascularity, while deep learning refines tumor-growth dynamics using patient PSA test time series and 3D physiological interactions. Real-patient reconstructions over 2.5 years achieve tumor-volume relative errors from 0.8% to 12.28%, including growth cases without PSA rises, supporting personalized monitoring protocols.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Physics-informed machine learning digital twin for reconstructing prostate cancer tumor growth via PSA tests

Motivation and clinical problem

Current diagnostic and monitoring approaches

Computational modeling prior work

Proposed digital-twin framework

npj | digital medicine Article | Published in partnership with Seoul National University Bundang Hospital | | | --- | --- | | [https://doi.org/10.1038/s41746-025-01890-x](https://doi.org/10.1038/s41746-025-01890-x) | | | Physics-informed machine learning digital twin for reconstructing prostate cancer tumor growth via PSA tests
Check for updates | | Daniel Camacho-Gomez1,2, Carlos Borau 1,3, Jose Manuel Garcia-Aznar1, Maria Jose Gomez-Benito1, Mark Girolami2,4 & Maria Angeles Perez1 Existing prostate cancer monitoring methods, reliant on prostate-speciﬁc antigen (PSA) measurements in blood tests often fail to detect tumor growth. We develop a computational framework to reconstruct tumor growth from the PSA integrating physics-based modeling and machine learning in digital twins. The physics-based model considers PSA secretion and ﬂux from tissue to blood, depending on local vascularity. This model is enhanced by deep learning, which regulates tumor growth dynamics through the patient’s PSA blood tests and 3D spatial interactions of physiological variables of the digital twin. We showcase our framework by reconstructing tumor growth in real patients over 2.5 years from diagnosis, with tumor volume relative errors ranging from 0.8% to 12.28% . Additionally, our results reveal scenarios of tumor growth despite no signiﬁcant rise in PSA levels. Therefore, our framework serves as a promising tool for prostate cancer monitoring, supporting the advancement of personalized monitoring protocols. Prostate cancer(PCa)is oneofthe most prevalent forms ofcancer that affect men1. It is estimated that prostate cancer accounted for1.4million new cases globally and resulted in more than 370,000 deaths in 2020 alone2. Prostate cancer is characterized by the uncontrolled growth and division of luminal cells within the prostate gland. Over time, these cancerous cells can invade nearby tissues and potentially spread to other parts of the body, mainly to the bones, lymph nodes, liver, and lungs3–6 in a process called metastasis7. Therefore, predicting the evolution of prostate cancer is essential for timely detection of growth and halting the expansion of the disease. The diagnosis of prostate cancer typically relies on the Prostate Imaging Reporting and Data System(PI-RADS), which assigns scoreson aﬁvepoint scale to lesions observed in magnetic resonance imaging (MRI) images8, and the Gleason score, which assesses the differentiation of cells in biopsy samples9, serving as an indicator of tumor prognosis10. After conﬁrming tumor presence through MRI and biopsy, prostate cancer monitoring is commonly based on the prostate-speciﬁc antigen (PSA) biomarker11. The PSA is a protein produced by both normal and cancerous cells within the prostate gland. Its main function is to liquefy semen, aiding in the mobility and transportation of sperm during ejaculation12. PSA levels in the blood have been extensively employed as a biomarker for both the detection and ongoing monitoring ofprostate cancer. Elevated levels ofPSA may indicate various prostate conditions, including prostate cancer13. While the exact reason for increased PSA levels in prostate cancer is not fully understood, it is believed that cancerous cells can disrupt the normal architecture of the prostate gland, leading to increased production and leakage ofPSA into thebloodstream. Consequently, measuring PSA levels in the blood can help in both the detection and the monitoring of prostate cancer. Yet, tumor progression and growth often occur without asigniﬁcant rise in PSA levels, thereby obscuring the prognosis ofthe tumor. The limited speciﬁcity and sensitivity of the PSA as biomarker14–16 are linked with poor diagnosis, as well as treatment and screening-related adverse effects17. Consequently, it is essential to deepen our understanding ofthe relationship between PSA levels and tumor development. Signiﬁcant improvements have been made in the comprehension of prostate canc

FAQs

Why can PSA-based monitoring fail to detect prostate tumor growth?

Tumor progression and growth can occur without a significant PSA rise, which reduces PSA sensitivity and specificity as a biomarker for prognosis.

How does the proposed digital-twin framework reconstruct tumor growth?

It combines a physics-based model of PSA secretion and tissue-to-blood flux tied to local vascularity with deep learning that regulates tumor growth dynamics using patient PSA blood tests and 3D interactions of physiological variables.

What performance did the framework achieve in real patient reconstructions?

Across reconstructions over 2.5 years from diagnosis, tumor-volume relative errors ranged from 0.8% to 12.28%.

Does the framework capture tumor growth even when PSA levels do not rise?

Yes. The results include scenarios of tumor growth despite no significant increase in PSA levels, improving monitoring beyond PSA-only trends.

Aug 5, 2026

Upload by Quinn