1. Efficient colorectal polyp segmentation using wavelet transformation and AdaptUNet: A hybrid U-Net.
- Author
-
Rajasekar D, Theja G, Prusty MR, and Chinara S
- Abstract
The prevalence of colorectal cancer, primarily emerging from polyps, underscores the importance of their early detection in colonoscopy images. Due to the inherent complexity and variability of polyp appearances, the task stands difficult despite recent advances in medical technology. To tackle these challenges, a deep learning model featuring a customized U-Net architecture, AdaptUNet is proposed. Attention mechanisms and skip connections facilitate the effective combination of low-level details and high-level contextual information for accurate polyp segmentation. Further, wavelet transformations are used to extract useful features overlooked in conventional image processing. The model achieves benchmark results with a Dice coefficient of 0.9104, an Intersection over Union (IoU) coefficient of 0.8368, and a Balanced Accuracy of 0.9880 on the CVC-300 dataset. Additionally, it shows exceptional performance on other datasets, including Kvasir-SEG and Etis-LaribDB. Training was performed using the Hyper Kvasir segmented images dataset, further evidencing the model's ability to handle diverse data inputs. The proposed method offers a comprehensive and efficient implementation for polyp detection without compromising performance, thus promising an improved precision and reduction in manual labour for colorectal polyp detection., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2024 The Authors.)
- Published
- 2024
- Full Text
- View/download PDF