Year: 2023
Type: article
Abstract: In this paper a high-resolution U-Net architecture for the segmentation of liver Computed Tomography (CT) scan images with high accuracy is presented. The contraction and expansion paths present in the proposed U-Net allow for reducing computational time and increase the resolution of information of the liver CT scan image respectively. The high-resolution U-Net architecture has been trained and validated using an IRCAD 2D liver CT image dataset. Performance of the U-Net architecture has been verified by calculating performance metrics of segmented 2D liver CT scan images. Results of performance metrics show that the proposed U-Net architecture achieves the best quality performance. The maximum dice coefficient of the liver segmentation during training phase is 95.79 % whereas during validation phase is 89.27 %. The high-resolution U-Net architecture has also compared with other researcher’s work in this paper.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7218272294
