Leveraging Machine Learning Algorithms to Forecast the Development of Prostate Cancer

Abstract

Prostate cancer is a leading cause of cancer-related deaths among men, with early detection crucial for improving outcomes. Traditional diagnostic methods like digital rectal exams (DRE), prostate-specific antigen (PSA) tests, and biopsies are invasive and imprecise. This study explores machine learning and deep learning techniques to predict prostate cancer development using clinical and imaging data. We compare classical machine learning models such as decision trees (DT), logistic regression (LR), and K-nearest neighbours (KNN), along with convolutional neural networks (CNNs), to improve diagnostic accuracy. A data preprocessing pipeline addressed class imbalance, feature selection, and model training. The dataset of 100 prostate cancer records was split into training and testing sets. KNN outperformed other models, achieving 92% accuracy, 0.88 precision, and 0.91 recall for malignancy. The DT model achieved 77% accuracy, while LR showed poor performance, especially for minority classes. Validation with 5-fold cross-validation confirmed KNN’s consistent performance with high precision and recall. The DT model offered interpretable results useful for clinicians. Especially KNN, shows promise in early prostate cancer detection and could support personalized care, addressing limitations of traditional diagnostic methods.

Ähnliche Arbeiten