A Comprehensive Survey of Image Segmentation for Medical Images

Medical imaging is the field in which doctors take the images of different interior body parts with the help of different techniques. This helps the medical professionals to do the diagnosis and treatment. Medical image analysis enhances the results of clinical research and it also helps to give better treatment solutions to the patients. Medical image processing includes tasks such as registration, segmentation, feature extraction, and classification. Segmentation of the medical images is a critical stage in the medical images processing, particularly as a necessary precondition for effective diseases diagnosis and therapy. Deep learning has gained popularity as a method for segmenting images. U-Net and its variations are currently the most extensively used method. Additionally, transformer-based models like TransUNet have given remarkable performance on different medical image datasets, with the higher success of pre-trained models in natural language processing field. The other hand these techniques have undergone some drawbacks. Considering CNN, It focuses only on local data and ignores global data. U Net networks are slow to utilize, parameter-heavy, and computationally demanding. The Transformer-based model has significant computational overhead. We have surveyed by summarizing segmentation methods in medical imaging. Additionally, we talked about well-liked segmentation strategies as they would be helpful to researchers in the future.