A Novel Liver Image Classification Method Using Perceptual Hash-Based Convolutional Neural Network

Abstract

Classification of liver masses plays an important role in early diagnosis of patients. This paper proposes a method to reduce the liver computed tomography (CT) images classification time and maintain the classification performance above an acceptable threshold by using convolutional neural network (CNN). A hybrid model called fused perceptual hash-based CNN (F-PH-CNN) is proposed by using a perceptual hash function together with the CNN. The proposed method has been designed for differential diagnosis between benign and malignant masses using CT images. The most important feature of the perceptual hash functions is to obtain the salient features of images. In the proposed F-PH-CNN method, DWT-SVD-based perceptual hash functions are used. The study uses CT images of 41 benign and 34 malign samples obtained from Elazig Education and Research Hospital. These samples were augmented up to 112 samples. The experimental results show that the CNN features achieved a better classification performance in which the ANN simulation results validate that the all output data with 98.2% success. The proposed method might also address the clinical computer-aided diagnosis of liver masses.