Comparison Analysis of Deep Learning Models In Medical Image Segmentation
Shivangi Tripathi, Rajesh Wadhwani, Akhtar Rasool, Abhishek Jadhav · 2023
Despite U-Net and its variants' notable accomplishments in the field of medical picture segmentation work, its segmentation capabilities intended for little matters are still not good enough. Segmentation in medical imaging has seen widespread use of deep learning. The benefits of precise segmentation of tiny targets and its scalable network design are demonstrated by U-Net. U-Net has received more than 2500 academic citations as a result of the recent increases in the performance requirements for segmentation in medical imaging. The U-Net architecture has been continuously developed by several academics. This paper reviews and categorises the related methodology, summarizes the medical image segmentation machineries based on the U-Net erection alternates in terms of the structure, origination, adeptness, etc., and introduces the loss functions, assessment constraints, and elements typically used for segmentation in medical imaging, which will serve as a useful guide for imminent study. Even though U-Net has confirmed reliable recital in semantic separation, it has been confirmed that U-Net has limits when faced with situations involving more image characteristics and semantic data. To address this, a multistate attention-thick U-net [3] has been created. Here, a comparative analysis of these deep- learning neural network architectures for image processing is done. The detailed comparison of U net and its variants is listed and explained.