Improving the Diagnostic of Contrast Enhanced Ultrasound Imaging using Optical Flow for Focal Liver Lesion Detection

Our proposal aims an automatic method used for obtaining the ultrasound image of a region of interest based on the optical flow computation. Combined with a kernel correlation filter tracking algorithm and a Xception deep convolutional neural network architecture, our solution provides state-of-the-art results (over 90% accuracy) in the automatic diagnosis of liver lesion using contrast enhanced ultrasound.