Related Experiment Video
Updated: Aug 2, 2026
04:57
Establishing a Competing Risk Regression Nomogram Model for Survival Data
Published on: October 23, 2020
10.1K
Machine Learning-based Nomograms for Predicting Clinical Stages of Initial Prostate Cancer: A Multicenter
Luyao Chen1, Zhehong Fu2, Qianxi Dong1
1Jiangxi Provincial Key Laboratory of Urinary System Diseases, Department of Urology, The First Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, China.
Urology
|August 17, 2024
Summary
Machine learning nomograms accurately predict prostate cancer (PCa) progression stages, aiding personalized treatment. These models, using biomarkers and clinicopathologic data, assist clinicians in categorizing patients for better prognosis assessment.
Area of Science:
- Oncology
- Medical Informatics
- Biostatistics
Background:
- Prostate cancer (PCa) staging is crucial for treatment and prognosis.
- Accurate prediction of advanced prostate cancer (APC) and metastatic prostate cancer (mPCa) is essential for patient management.
- Existing methods for PCa progression prediction may benefit from advanced computational approaches.
Purpose of the Study:
- To develop and validate machine learning (ML)-based nomograms for predicting PCa progression stages.
- To identify optimal biomarkers and clinicopathologic features for distinguishing between different PCa progression stages.
- To assess the clinical utility of ML-derived nomograms in aiding treatment decisions and prognosis.
Main Methods:
- Construction of training, testing, and external validation sets from PCa patient cohorts.
- Identification of significant predictors for APC and mPCa using logistic regression (LR).
- Development and comparison of various ML models, with the best-performing algorithm used for nomogram creation.
- Evaluation of nomogram performance using AUC, calibration curves, and decision curve analysis (DCA).
Main Results:
- Logistic regression identified PSA, maximum tumor diameter, Gleason score, and RNF41 as key predictors for APC.
- PSA, ALP, and Gleason score were significant for predicting mPCa.
- LR-based nomograms demonstrated superior predictive performance across all sets, with high AUC values (e.g., 0.848 for APC, 0.940 for mPCa in training sets).
- Calibration and DCA confirmed the nomograms' reliability and clinical usefulness.
Conclusions:
- Machine learning-based nomograms, particularly those derived from LR, show excellent predictive accuracy for PCa progression.
- These validated nomograms can effectively assist clinicians in stratifying patients with newly diagnosed PCa.
- The developed nomograms support personalized treatment planning and enhance prognosis assessment for prostate cancer patients.
