Empowering Liver Disease Diagnosis with Convolutional Neural Networks

Hepatic diseases are responsible for 4% of total deaths, which accounts to roughly two million deaths per year, of which two-thirds are liver related deaths that occur in men. Liver inflammation can lead to Hepatitis condition that results in an infection, this causes scarring of liver a condition known as Liver Cirrhosis. In our work a commonly used Noninvasive scan, Ultrasound was used to acquire 152 images across Healthy, Fatty, Fibrosis and Cirrhosis Liver. The primary purpose of this paper is to use the convolution neural net models that demonstrate high sensitivity and accuracy in classifying liver disease stages. Best CNN model resulted in a sensitivity of 90% for normal and 85.71% for abnormal ultrasound liver images, this would help the clinician to reduce the observational bias and make faster decisions in diagnosing the patients with liver diseases.