Harnessing machine learning to predict prostate cancer survival: a review

Abstract

The prediction of survival outcomes is a key factor in making decisions for prostate cancer (PCa) treatment. Advances in computer-based technologies have increased the role of machine learning (ML) methods in predicting cancer prognosis. Due to the various effective treatments available for each non-linear landscape of PCa, the integration of ML can help offer tailored treatment strategies and precision medicine approaches, thus improving survival in patients with PCa. There has been an upsurge of studies utilizing ML to predict the survival of these patients using complex datasets, including patient and tumor features, radiographic data, and population-based databases. This review aims to explore the evolving role of ML in predicting survival outcomes associated with PCa. Specifically, we will focus on the applications of ML in forecasting biochemical recurrence-free, progression to castration-resistance-free, metastasis-free, and overall survivals. Additionally, we will suggest areas in need of further research in the future to enhance the utility of ML for a more clinically-utilizable PCa prognosis prediction and treatment optimization.

Keywords: artificial intelligence; machine learning; precision medicine; prostate cancer; survival.

Figure 1

Harnessing machine learning to predict prostate cancer survival: a comprehensive review. ANN, Artificial Neural Networks; kNN, k-Nearest Neighbor; RF, Random Forest Classifier; NBC, Naïve Bayes Classifier; SVM, Support Vector Machine; CNN, Convoluted Neural Network; XGBooost, Extreme Gradient Boosting; GBM, Gradient Boosting Machine; DT, Decision Tree; MLP, Multilayer Perceptron; LSTM, Long Short-Term Memory.