Automated Detection and Classification of Liver Cancer from CT Images using HOG-SVM model

Liver cancer patients have a high death rate due to the diagnosis of the disease in the final stages. Computer-aided diagnosis from various medical imaging techniques can assist significantly in detecting liver cancer at a very early stage. This paper presents an automated method of detecting liver cancer in abdominal CT images and classifying them using the histogram of oriented gradient - support vector machine (HOG-SVM) algorithm. The proposed model consists of several stages where the image is first normalized and preprocessed using a Median and Gaussian filter to remove noise in the image. The image segmentation and liver area extraction are executed in the second stage combining thresholding and contouring. We integrated an ROI based histogram oriented gradient (HOG) feature extraction to train the classifier which impels the classification faster than the conventional methods. Finally, liver CT images are classified implementing support vector machine and segmented results are highlighted with different markers. The proposed system is tested on real data of 27 confirmed early-stage liver cancer and the experimental result shows an accuracy of 94% detecting liver cancer.