Computed Tomography Visual Question Answering with Cross-modal Feature Graphing