Machine Learning based Early Stage Identification of Liver Tumor using Ultrasound Images

Abstract

Liver cancer is one of the most malignant diseases and its diagnosis requires more computational time. It can be minimized by applying a Machine learning algorithm for the diagnosis of cancer. The existing machine learning technique uses only the color-based methods to classify images which are not efficient. So, it is proposed to use texture-based classification for diagnosis. The input image is resized and pre-processed by Gaussian filters. The features are extracted by applying Gray level co-occurrence matrix (GLCM) and Local binary pattern (LBP in the preprocessed image. The Local Binary Pattern (LBP) is an efficient texture operator which labels the pixels of an image by thresholding the neighborhood of each pixel and considers the result as a binary number. The extracted features are classified by multi-support vector machine (Multi SVM) and K-Nearest Neighbor (K-NN) algorithms. The Advantage of combining SVM with KNN is that SVM measures a large number of values whereas KNN accurately measures point values. The results obtained from the proposed techniques achieved high precision, accuracy, sensitivity and specificity than the existing method.