Ultrasonic diagnosis of cirrhosis based on preprocessing using pyramid recurrent neural network

In this paper, a pyramid recurrent neural network is applied to characterize the hepatic parenchymal diseases in ultrasonic B-scan texture. The cirrhotic parenchymal diseases are classified into four types according to the size of hypoechoic nodular lesions. The B-mode patterns are wavelet transformed, and then the compressed data are fed into a pyramid neural network to diagnose the type of cirrhotic diseases. Compared with the three-layer neural networks, the performance of the proposed pyramid recurrent neural network is improved by a more efficient utilization of lower layers. Simulation results show that the proposed system is suitable for diagnosis of cirrhosis diseases.