We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Towards improved U-Net for efficient skin lesion segmentation.
- Authors
Nampalle, Kishore Babu; Pundhir, Anshul; Jupudi, Pushpamanjari Ramesh; Raman, Balasubramanian
- Abstract
Skin cancer is a highly lethal disease, and detecting it at an early stage is critical. Skin lesion segmentation is a complex process involving identifying the infected area in an image with low contrast, variable size, and position. This task is essential in medical analysis, as it helps clinicians focus on a specific area of the image before further analysis. Our paper introduces a new method for improving the segmentation of medical images by providing the efficient neural connections to design efficient U-Net architecture. We have utilized skip paths to the encoder and minimize the semantic gap between concatenated feature maps. This leads to more precise segmentation outcomes. We have used the PH2 and ISIC-2018 as benchmark dataset to validate the effectiveness of the proposed approach and surpass the available benhcmark performance. We have obtained approximately 96.18% accuracy with the PH2 dataset and 96.09% accuracy with the ISIC-2018 dataset. The outcomes of our architecture are quite impressive, and they exhibit superior performance over both the baseline model and other state-of-the-art techniques.
- Subjects
SKIN imaging; DEEP learning; SKIN cancer; DIAGNOSTIC imaging; MEDICAL personnel
- Publication
Multimedia Tools & Applications, 2024, Vol 83, Issue 28, p71665
- ISSN
1380-7501
- Publication type
Article
- DOI
10.1007/s11042-024-18334-5