The integration of artificial intelligence (AI) into medical image analysis has revolutionized diagnostic practices, offering significant improvements in accuracy and efficiency. This article presents a comprehensive review of both foundational concepts and advanced approaches in the processing and analysis of medical images, with a focus on segmentation and classification tasks. We explore the evolution from traditional image processing techniques to machine learning (ML) and deep learning (DL) methods, such as support vector machines (SVMs), random forests, and convolutional neural networks (CNNs) and his approaches, which have demonstrated strong performance in delineating regions of interest and distinguishing between pathological categories. Particular attention is given to the role of these technologies in the diagnosis of diseases like prostate and lung cancer, and other diseases. Despite their success, several challenges persist, including the generalization of models across heterogeneous datasets, variability in anatomy and imaging conditions, and the interpretability of results. Additionally, computational complexity and the misuse of evaluation metrics in imbalanced data scenarios limit clinical applicability. This review discusses these issues and outlines future directions, such as the development of resource-efficient architectures, improved evaluation protocols, and hybrid optimization methods to enhance model performance and transparency. These advances are crucial to bridging the gap between research and clinical deployment, contributing to more accurate, explainable, and personalized medical care.