Prostate gland cancer detection and grade classifying using gleason score-based machine learning approach

Detection and grading of the prostate cancer can only be effective when the clinical decision-making is done accurately, but the traditional Gleason scoring is subjective in nature and is subject to inter observers variation. This paper will solve these problems by proposing a machine learning-based model of automated prostate cancer detection and Gleason score classification (210) with the help of histopathologic images and clinically significant biomarkers. The suggested method combines the most modern feature extraction features, such as texture analysis (e.g., Haralick features) and morphological characterization to form a discriminative feature space. Various supervised learning structures such as Support Vector Machines (SVM), Random Forest (RF), and deep learning models with Convolutional Neural Networks (CNNs) are used in the classification. In addition, a superior hybrid framework is discussed to ensure that the local spatial variation and global contextual variation are captured to ensure that grading accuracy is improved. The sample size used to train and test the model is 3000 with data augmentation to enhance generalization. The experimental findings indicate that the CNN-based model has better performance and reaches an accuracy of 94.5, precision of 93.8, recall of 94.2, and F1-score of 94.0, which is better than the traditional machine learning methods. The suggested framework offers a credible and scalable decision-support system that minimizes variations in diagnoses, improves consistency in grading, and aids timely and accurate diagnosis of prostate cancer, which further provides a positive contribution to the clinical outcomes.