Abstract
Liver tumour segmentation is a challenging task due to the wide diversity in size, position, depth, and proximity to surrounding organs. This research uses the state-of-the-art model of Mask R-CNN model with the ResNet-50 architecture as the backbone. The suggested methodology leverages the Mask Region-Convolutional Neural Network approach to accurately identify liver tumors by identifying tumour location. To address variations of the liver and CT scan images with different parameters. The normalized CT images are then fed into the RESNET-50 model to extract relevant features. Subsequently, the liver tumor are segmented using the Mask R-CNN algorithm. The experimental dataset used in this study consists of one hundred and thirty CT scans obtained from various hospitals and nursing homes, which are freely accessible on the LiTS web page. The suggested algorithm is trained on transformed CT image slices. The results demonstrate that the proposed Mask RCNN system, with its innovative connections, surpasses state-of-the-art methods in identifying liver tumor, achieving a remarkable DSC value of 0.97%. This technique has the potential to significantly contribute to early and precise diagnosis of liver tumor in the field of biotechnology, potentially saving many patients' lives.