A non-invasive approach to prostate cancer diagnosis with MRI-based feature extraction and machine learning

Prostate cancer, a significant global health concern, requires accurate and timely diagnosis for optimal treatment planning. Gleason grading, a key prognostic factor, often relies on subjective histopathological interpretation. This research proposes a non-invasive approach to accurately predict prostate cancer Gleason grade using Magnetic Resonance Imaging (MRI) and advanced machine learning techniques. Our method involves three key steps: image segmentation, feature extraction, and classification. First, we utilize Fuzzy C-Means segmentation to accurately identify tumor regions within MRI scans. Subsequently, a modified ResNet50 model is employed to extract relevant features from these segmented regions. Finally, these extracted features are fed into a machine learning models classifiers to predict the Gleason grade of the tumor Through this experimentation, we found that the Gradient Boosting Machine (GBM) model consistently outperforms other popular techniques like Random Forest and Support Vector Machine (SVM). The GBM model achieved an impressive accuracy of 93%, sensitivity of 92%, specificity of 94%, and an AUC-ROC of 0.95. These impressive metrics indicate that our GBM model can effectively distinguish between low-grade and high-grade prostate cancers. By automating the diagnostic process and reducing reliance on subjective human interpretation, our proposed method has the potential to significantly improve patient outcomes and treatment decisions. This innovative approach could lead to earlier and more accurate diagnosis of prostate cancer, enabling timely and effective treatment interventions.