A Performance Comparison on the Machine Learning Classifiers in Predictive Pathology Staging of Prostate Cancer

Area of Science:

• Urology
• Oncology
• Medical Informatics

Background:

• Accurate pathological staging is crucial for guiding treatment decisions in prostate cancer.
• The Partin table is a widely used tool for predicting pathological stage but may have limitations in specific populations.
• Evaluating advanced predictive models is essential for improving staging accuracy.

Purpose of the Study:

• To investigate the predictive performance of the Partin table and machine learning methods for pathological stage prediction.
• To assess the accuracy of these models using a dataset of Korean patients.
• To identify superior methods for pathological staging in this demographic.

Main Methods:

• Utilized the SPCDB dataset comprising records from 944 patients treated at a tertiary hospital.
• Applied the traditional Partin table for pathological stage prediction.
• Implemented and evaluated several machine learning algorithms, including Support Vector Machines (SVM).

Main Results:

• The Partin table demonstrated low accuracy (65.68%) when applied to the Korean dataset for patients with organ-confined and non-organ-confined conditions.
• Support Vector Machines (SVM) achieved a higher prediction accuracy of 75%.
• SVM emerged as a more promising alternative for pathological staging compared to the Partin table.

Conclusions:

• The Partin table's predictive accuracy is limited when applied to Korean patient data.
• Machine learning, specifically SVM, offers improved accuracy for pathological stage prediction in this population.
• These findings suggest that SVM can enhance pathology staging, potentially leading to more personalized treatment strategies.