A Comparative Study on Liver Tumor Detection Using CT Images

Abstract

Liver cancer (LC) is a globally known issue. It is one of the most common cancers that can cause human beings. It is a fatal disease spreading especially in developing countries. Many algorithms have been used to perform the detection of liver cancer with the help of both traditional machine learning classifiers and deep learning classifiers. To analyze the performance of commonly used algorithms, this paper attempts a comparative study on LC detection. It includes both machine learning and deep learning techniques; and several methods for liver and tumor detection from CT images are used. With the advances in Artificial Intelligence (AI) and convolution neural networks algorithms, the methods included in this comparative study achieved great results. The best accuracy among traditional machine learning classifiers reaches 90.46% using Support Vector Machine (RBF). Inception V4 pre-trained model obtained 93.15% in terms of testing accuracy, and it is the best classifier among deep learning models. The performance of deep learning models is very promising to take place in medical decisions.