Predicting and Staging Hepatocellular Carcinoma from Contrast CT Scans

Abstract

Hepatocellular carcinoma (HCC) is a common and deadly form of liver cancer for which early detection and staging can be integral to patient survival. Medical imaging is an usual method of diagnosis, either using contrast Computed Tomography (CT) scans or Magnetic Resonance Imaging (MRI) scans. We introduce a new deep learning model that aims to take advantage of the information in two different stages of contrast CT scans to predict the presence and severity of HCC tumours in the images. Our model is trained and tested on a dataset of 307 labelled dual image input slices. On testing, the model achieves an accuracy of 96.8 % and a sensitivity of 87.8 %. These results indicate that using a dual image input of contrast CT scans provides a significant boost in performance to the model. Such a model prove to be a valuable tool to assist doctors in the diagnosis and staging of HCC, saving them time in the manual examination of scans. Implementation is publicly available at https://github.com/ZakirANU/CNN4LiverCancer. Show more