Work

Radiomics-Based Prediction of Bone Metastases from Baseline Prostate MRI in Prostate Cancer: A Systematic Review and Critical Appraisal

Year: 2026

Type: article

Abstract: Introduction: Whether bone metastasis is present at diagnosis determines curative versus systemic treatment of prostate cancer, yet referral for advanced staging relies on imperfect risk stratification; magnetic resonance imaging (MRI) radiomics has been proposed to predict metastatic status from routinely acquired staging imaging. Methods: This systematic review (PROSPERO CRD420261463568) searched PubMed, Embase and Cochrane CENTRAL for studies of radiomic, deep-learning or hybrid models predicting bone-metastatic status from prostate MRI; incremental value over clinical predictors, reporting completeness, radiomics methodology and risk of bias were appraised. Results: Six retrospective studies (1240 patients, 491 with bone metastasis, all from China; 322 contributed to the principal validation estimates) were included. Clinical-plus-radiomics models achieved validation areas under the receiver operating characteristic curve (AUCs) of 0.893–0.920 internally and 0.903 externally, with the highest estimate (0.928) from a multimodal model lacking clinical predictors. Only two studies compared clinical-only with combined models (AUC increments of 0.060 and 0.081). Conclusions: Reported discrimination was generally high, but current evidence cannot establish that radiomics adds value beyond routine clinical predictors. These models are best understood as predictive biomarkers rather than detection tools; a potential triage role among patients below routine staging thresholds remains untested, and transportability beyond the high-prevalence cohorts studied is unproven; prospective multicentre validation is required before clinical use.

Source: BioTech

DOI: 10.3390/biotech15050102

Authors: Nicholas A. Clausen, Hugo C. Temperley, Cormac B. Mulhall, Benjamin M. Mac Curtain, Louise C. McLoughlin, Michael E. Kelly, Ardeshir R. Rastinehad

Institutions: St John of God Murdoch Hospital, Trinity College Dublin, St. James's Hospital, The University of Western Australia, National University of Ireland, Ollscoil na Gaillimhe – University of Galway, Royal Perth Hospital, Northwell Health

Language: English

Topic: Radiomics and Machine Learning in Medical Imaging

Subfield: Radiology, Nuclear Medicine and Imaging

Field: Medicine

Domain: Health Sciences

Open Access status: gold

Estimated APC paid: $1,930

OpenAlex ID: W7225620094

FWCI:

Cites:

Cited by:

Related to: