Automated liver segmentation from CT images using modified ResUNet

The Segmentation of liver from computed tomography (CT) images is a critical task in medical research particularly when employing deep learning methods. Most doctors prefer to use Computed tomography images for liver disease identification hence automatic liver segmentation plays a decisive role. To perform automatic segmentation of liver popular architectures such as modified ResUNet are specifically designed to capture multi-resolution features. In this study we proposed an automatic system that utilizes convolutional layers to efficiently extract features while maintaining spatial information. A modified Residual UNet model was developed for liver segmentation using the 3Dircadb dataset, the model demonstrated excellent performance by achieving a Dice Score of 93.08% , accuracy of 98.57%, Jaccard Index of 87.12%, a volumetric overlap error of 12.87 % , and a relative volume difference of 14.91% on 128x128 size images. These results highlight the effectiveness of the proposed method. The primary goal of this research is to apply deep learning techniques for liver segmentation and assess the model’s performance using segmentation metrics.