Convolutional neural network classification of ultrasound images by liver fibrosis stages based on echo-envelope statistics

Introduction: Assessing the stage of liver fibrosis during diagnosis and follow-up patients with diffuse disease is crucial. The tissue structure in fibrotic reflected texture contrast an ultrasound image, pixel brightness indicating intensity echo envelope. Therefore, progression can be evaluated non-invasively by analyzing images. Methods: A convolutional-neural-network (CNN) classification images was applied to estimate fibrosis. In this study, colorization using echo-envelope statistics that correspond features proposed improve accuracy CNN classification. method, image modulated 3rd- 4th-order moments brightness. two original were then synthesized into a color RGB representation. Results Discussion: colorized classified via transfer learning VGG-16 evaluate effect colorization. Of 80 stages F1–F4, 38 accurately images, whereas 47 method.