Automated Detection of Primary Liver Cancer Using Different Deep Learning Approaches

Document Type : Original Article


1 Biomedical Engineering, Faculty of Engineering, Minia University, Minia

2 Electrical Engineering Department - Faculty of Engineering- Minia University.

3 Biomedical Engineering Department - Faculty of Engineering- Minia University.


Hepatocellular carcinoma (HCC) is the world's sixth-most common malignancy and the fourth-leading cause of cancer death. This paper introduces an early diagnose to HCC using fully autonomous system by artificial intelligence (AI) techniques. This system consists of three stages: preprocessing, liver segmentation, and classifier stage. Preprocessing improves the quality of the abdominal Computed Tomography(CT) scan to be better analyzed. The liver is extracted and segmented from improving CT image using various deep learning techniques, and semantic segmentation is used to separate lesions from a tumor-filled liver in cases of HCC by classifier. U-Net and Resnet50 deep learning algorithms are used to segment the liver and tumour automatically, but the Resnet50 achieved the better results with a DICE score of 0.926, and loss of 0.0028. The classifier's robustness is justified using 5-fold cross-validation (CV). The Resnet 50 classifier, which is a classification approach based on transfer learning (TL), achieved 96% accuracy for the binary classifier. In comparison with the linked methods, the proposed system confirms that it is the promised fully autonomous system for liver segmentation and tumour segmentation in cases diagnosed with HCC.


Main Subjects