Evaluation of a PSA and transrectal prostate ultrasound video-based machine learning model as a tool for prostate cancer diagnosis

Q: What is the main goal of the proposed model?

To build a machine learning model that uses PSA levels together with prostate ultrasound video clips to support prostate cancer diagnosis.

Q: How were the models trained and validated?

The study enrolled 928 participants and randomly selected 742 for model validation, while the remaining 186 were used to evaluate accuracy and reliability.

Q: Which algorithm performed best and how was interpretability assessed?

The XGBoost model achieved the most promising performance. SHAP visualization was used to identify the most influential features driving predictions.

Aug 3, 2026

Upload by Evangeline

Developing an accurate, non-invasive approach for prostate cancer diagnosis motivates machine learning models that integrate serum prostatespecific antigen (PSA) with prostate ultrasound video information. A cohort of 928 participants was used to train and assess models, combining ultrasound-derived video clip features with PSA and clinical indicators. Six algorithms were compared using ROC-based performance, while SHAP was applied to interpret feature contributions. Results show strong test-set discriminative ability and highlight XGBoost with the most promising accuracy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Introduction

Clinical background and screening limitations

Rationale for radiomics and machine learning

Methods

Participants and validation strategy

Feature extraction and model development

ROC evaluation and SHAP interpretation

Results

Comparative model performance (AUC)

Feature importance in the best model

Conclusion

TYPE Original Research PUBLISHED 08 September 2025 DOI 10.3389/fonc.2025.1590396 OPEN ACCESS EDITED BY Taja Lozar, Institute of Oncology Ljubljana, Slovenia REVIEWED BY Jincao Yao, University of Chinese Academy of Sciences, China Xinrui Huang, Peking University, China *CORRESPONDENCE Yanhong Du [duyh520@outlook.com](duyh520@outlook.com)[ ](duyh520@outlook.com)Liyan Hu [hlyhzh@outlook.com](hlyhzh@outlook.com)[ ](hlyhzh@outlook.com)Xiaoyang Qi [qixiaoyang16@163.com](qixiaoyang16@163.com) RECEIVED 09 March 2025 ACCEPTED 19 August 2025 PUBLISHED 08 September 2025 CITATION Du Y, Zhao A, Zhang M, Wang Z, Hu L and Qi X (2025) Evaluation of a PSA and transrectal prostate ultrasound video-based machine learning model as a tool for prostate cancer diagnosis. Front. Oncol. 15:1590396 . doi: 10.3389/fonc.2025.1590396 COPYRIGHT © 2025 Du, Zhao, Zhang, Wang, Hu and Qi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Evaluation of a PSA and transrectal prostate ultrasound video-based machine learning model as a tool for prostate cancer diagnosis Yanhong Du*, Anli Zhao, Maoliang Zhang, Zhengping Wang, Liyan Hu* and Xiaoyang Qi* Department of Ultrasound, The Afﬁliated Dongyang Hospital of Wenzhou Medical University, Dongyang, Zhejiang, China Objective: To develop a machine learning-based model incorporating prostatespeciﬁc antigen (PSA) levels and prostate ultrasound video clips for diagnosing prostate cancer. Methods: The study enrolled 928 participants, of whom 429 had prostate cancer and 499 other non-prostate cancers. Univariate and multivariate analyses of serological indices were conducted to detect signiﬁcant variables. From this cohort, 742 patients were randomly chosen for model validation, while the other 186 were employed to evaluate the accuracy and reliability of the model. Seven features were extracted from ultrasound video clips and combined with PSA and other clinical indicators. Predictive models were established using six machine learning algorithms and receiver operating characteristic (ROC) curves were used to determine the optimal model. SHapley Additive exPlanations (SHAP) was utilized to visualize feature importance in the best-performing model. Results: All six of the evaluated machine learning models performed favorably, with area under the ROC curve (AUC) values in the test set ranging from 0.800 to 0.881. Of these models, the XGBoost model achieved the most promising performance, signiﬁcantly surpassing that of the other models (P < 0 . 05) . SHAP visualization revealed that PSA, prostatic volume(PV), age, wavelet. LHL.ﬁrstorder. Median, wavelet. HLH.glszm.ZoneEntropy, and original.shape.MinorAxisLength were the most inﬂuential features in the XGBoost model. Conclusion: The developed machine learning models demonstrated signiﬁcant potential for prostate cancer diagnosis. Among them, the XGBoost model outperformed the others, highlighting its superior predictive capability. KEYWORDS prostate cancer, PSA, machine learning, prostate ultrasound video, SHAP Frontiers in Oncology 01 [frontiersin.org](frontiersin.org) 1 Introduction Prostate cancer(PCa) ranks second in global male malignancies, after only lung cancer (1) . In China, both diagnoses and deaths associated with prostate cancer have been increasing steadily in recent years (2) . Per the World Health Organization, in 2020, China reported an incidence rate of 15.6 per 100,000 individuals, with more than 110,000 new diagnoses and more than 50,000 deaths, making it a signiﬁcant public health concer

FAQs

What is the main goal of the proposed model?

To build a machine learning model that uses PSA levels together with prostate ultrasound video clips to support prostate cancer diagnosis.

How were the models trained and validated?

The study enrolled 928 participants and randomly selected 742 for model validation, while the remaining 186 were used to evaluate accuracy and reliability.

Which algorithm performed best and how was interpretability assessed?

The XGBoost model achieved the most promising performance. SHAP visualization was used to identify the most influential features driving predictions.

Aug 3, 2026

Upload by Evangeline