Automated Classification of Hepatocellular Carcinoma (HCC) Images for Detection of Malignant Tumor Using HOG Technique

Abstract

AbstractPatients with liver cancer have a high mortality rate before the final diagnosis. Computerized medical diagnosis by medical imaging methods may assist significantly in disease diagnosis during the onset of cancer. This work presents an automated method of identifying liver cancer in humans using CT scans and classifying them with the histogram of oriented gradient with different paradigms of support vector machine. The model implements the image normalization and preprocessing using homomorphic and median filters to remove artifacts in the image. The segmentation and area extraction are done in the second step by thresholding and fitting the contour model. ROI-based gradient histogram is used to extract features to train the classifier with faster rate in fine Gaussian SVM relative to other SVM models. It is found that supervised learning method has better performance in accuracy and more effective in liver segmentation and identification of malignant tumor. The experimental results demonstrate that the model shows an accuracy of 98.4% detecting liver cancer in the real evidence.KeywordsComputer tomography (CT)Support vector machines (SVM)Histogram of oriented gradient (HOG)Region of interest (ROI)Feature extractionClassification