Auxiliary Medical Decision System for Prostate Cancer Based on Ensemble Method

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Prostate cancer (PCa) poses a significant global health challenge, particularly in resource-limited settings where patient-doctor conflicts arise.
• The need for efficient diagnostic tools is critical to manage the growing burden of PCa.
• Existing diagnostic methods may not fully leverage large-scale patient data for improved decision-making.

Purpose of the Study:

• To develop an auxiliary medical decision system for prostate cancer (PCa) utilizing machine learning.
• To leverage a large dataset of patient information for enhanced diagnostic capabilities.
• To analyze PCa incidence trends and the impact of lifestyle and genetics.

Main Methods:

• Utilized six tumor markers as input features for machine learning models.
• Employed support vector machine (SVM) and artificial neural network (ANN) models.
• Implemented a stacking ensemble method to mitigate overfitting and improve model robustness.
• Trained the models on a dataset of 1,933,535 patient records from three major hospitals over five years.

Main Results:

• The developed auxiliary medical decision system effectively processes large volumes of patient data.
• System performance demonstrated continuous improvement with increasing data volume.
• Statistical analysis revealed an increasing trend in prostate cancer incidence over the past five years.
• Identified high-fat diets and genetic inheritance as significant factors negatively impacting PCa prevalence.

Conclusions:

• The auxiliary medical decision system shows promise in supporting clinical decisions for prostate cancer.
• Massive data integration enhances the predictive power and utility of the system.
• Dietary habits and genetic predisposition are critical modifiable and non-modifiable risk factors for prostate cancer.