Liver Tumour Classification and Segmentation Through Machine Learning Techniques

Liver Tumour is the \(6\text {th}\) most common type of cancer worldwide, accounting for more than 800,000 deaths per year. The application of modern AI/ML-based technology has been growing in the area of medical science, which can help in an early and precise detection of the tumour. This research work begins by highlighting the work done so far in field of bio-medical images in terms of algorithms, challenges and steps to over come them. Based on the studies, this paper provides the approach for liver tumour segmentation on the CT scan records using the U-Net architecture as base. Further paper provides an approach of using ResNet50 as encoder of U-Net architecture for the same data set and comparing the results of the two approaches on the accuracy and dice coefficient metrices. 50 CT scans records available on Kaggle were used for this study.