Work

Diagnostic value of multiparametric MRI radiomics combined with serum PSA and tissue P504S for prostate cancer

Year: 2026

Type: article

Abstract: This study aimed to evaluate the diagnostic value of multiparametric MRI radiomics features combined with serum PSA-related indicators and tissue P504S expression for prostate cancer, and to develop a multimodal auxiliary diagnostic model in a retrospective cohort that was not restricted to early-stage disease. Clinical, imaging, and pathological data from 126 patients with suspected prostate cancer were retrospectively collected and divided into a prostate cancer group (69 cases) and a benign lesion group (57 cases). All patients underwent prostate MRI; radiomics features were extracted, and LASSO regression, with 10-fold cross-validation for tuning, was used to construct a Rad-score. Serum t-PSA, f-PSA, f/t-PSA ratio, and PSAD were measured. Tissue P504S expression was detected by immunohistochemistry and semi-quantitatively scored. A combined diagnostic model was constructed using multivariate logistic regression. ROC metrics were evaluated in the model-development cohort and are reported as apparent performance. AUC differences between models were compared using the DeLong test. The prostate cancer group had higher t-PSA and PSAD and a lower f/t-PSA ratio than the benign lesion group (all P < .001). P504S was positive in 64/69 cancer cases (92.8%) and 9/57 benign cases (15.8%; P < .001). Eight radiomics features were selected, and the median Rad-score was 0.86 in cancer cases and - 0.52 in benign cases (P < .001). The full model combining Rad-score, f/t-PSA, PSAD, and tissue P504S semiquantitative score achieved an apparent AUC of 0.956 (95% CI: 0.918-0.994), sensitivity of 92.8%, specificity of 91.2%, PPV of 92.8%, and NPV of 91.2%. The overall imaging + serum model had an apparent AUC of 0.912 (95% CI: 0.856-0.968). In exploratory location-based analyses, AUCs were 0.962 for peripheral zone lesions and 0.934 for transition zone lesions. The full model showed promising apparent discrimination for prostate cancer in this retrospective cohort and may represent a potential auxiliary diagnostic model after tissue sampling. The imaging + serum model could potentially support biopsy decision-making, whereas the full model requires tissue P504S assessment and may assist diagnostic confirmation or risk characterization. Neither incremental clinical benefit nor a reduction in unnecessary biopsies has been established. Internal and external validation and evaluation of the practical clinical workflow are required before routine use.

Source: Medicine

DOI: 10.1097/md.0000000000051391

Authors: Lu Zhang, Deyuan Zhao, Yujing Bai, Dengcui Zhao, Jian Li

Institutions: Hebei Medical University, First Affiliated Hospital of Hebei Medical University

Language: English

Topic: Radiomics and Machine Learning in Medical Imaging

Subfield: Radiology, Nuclear Medicine and Imaging

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $1,950