Since ancient times in in medical imaging, image segmentation plays an important role, it widely used for object recognition. This paper provides a comprehensive exploration of image segmentation techniques, analyzing their theoretical foundations, practical implementations, and advantages across medical domain. We start our overview of traditional segmentation approaches, including thresholding, edge detection, and region-based methods, establishing a strong conceptual framework. Building upon this foundation, using this as a starting point, we will gradually look at more complex methods, especially new ways to separate things using artificial intelligence. Deep learning-based methods, including fully convolutional neural networks (CNNs), U-Net, and vision transformers (ViTs), are explored in detail, highlighting their ability to reach the best possible results. This paper combines old methods with new improvements to provide a complete view of image cutting and its various applications in medical pictures.