Machine Learning Methods for Prostate Cancer Prediction using Magnetic Resonance Imaging and Clinical Data

Abstract

Prostate cancer is one of the most common cancers in males and is difficult to diagnose in its early stages. Traditional methods have attempted to predict prostate cancer from clinical data, but lack of ability to incorporate image to enhance the predictive outcomes. To address this issue, we develop an analysis pipeline that integrates clinical and medical imaging data to efficiently predict prostate cancer using multiple machine learning approaches. In this paper, we demonstrate a high performance of our analysis pipeline by using Magnetic Resonance Imaging, clinical data and advanced pre-processing technique to overcome class imbalance challenges. In the task of analyzing prostate cancer datasets from Challemon Challenge, our analysis pipeline delivers a high accuracy score of 86.9% in prediction using deep neural network architecture, showing strong superiority over traditional methods. This research is not only promising in providing improvement in patient outcomes but also optimizing use of healthcare resources.