Automatic Detection of Focal Liver Lesions in Multi-phase CT Images Using A Multi-channel & Multi-scale CNN

There are multiple types of tumors occurring in the liver, each of which have a different visual appearance that changes after injection of the contrast medium. Therefore, detection of liver tumors is considered to be a challenging task. In the present article, the authors propose a method for detection of liver tumor candidates with different size in multi-phase CT images using a multi-channel and multi-scale CNN protocol. Experimental results demonstrated that recall scores and dice coefficient can be significantly improved, with a reduction in the number of average false positives using the proposed method compared with existing methods.