U-Net Based Image Segmentation Drawbacks in Medical Images: A Review

In the modern healthcare segmentation of the medical image plays a crucial role which lets the accurate delineation of anatomical structures for medical diagnosis and treatment. U-Net is an emerging convolutional neural network which is good in medical image processing. The U-Net model was developed by Ranneberger, Fischer and Brox in 2015. This work is highlighting the inherent drawbacks and challenges with U-Net based image processing specially in medical imaging. Since U-Net has proved itself in different scenarios, however it has some challenges like susceptibility to imbalance class, and limiting contextual data and sensitivity to variations in image quality pose significant obstacles. These challenges need to be addressed to ensure reliable and clinically relevant results in medical image segmentation. The work is exploring the recent advancements and reviewing these drawbacks in depth and discussing their impacts on segmentation, specificity and sensitivity.