Determination of Accuracy of Neural Network Method Using Magnetic Resonance Images in Finding Liver Cancer Level

Abstract

This paper proposes the detection of lever cancer by image segmentation via Convolutional Neural Network and comparing accuracy and sensitivity with K-Nearest Neighbor Classifier. 40 samples have been considered for this work. Convolutional Neural Network contains 20 samples in group 1 and group 2 has 20 samples for K-Nearest Neighbor Classifier. With a pretest power of 80%, an independent sample T-test were performed for both the groups. An accuracy of 96.29% is achieved by Convolutional Neural Network and K-Nearest Neighbor achieves an accuracy of 89.96% with significance of p<0.05. The Sensitivity of 97.61% and 95.38% with significance of p<0.05 is achieved by convolutional Neural Network and K-Nearest Neighbor respectively. Convolutional Neural Network accomplishescomparatively better sensitivity and accuracy in cancer segmentation of liver when compared with K-Nearest Neighbor classifier.