Mia Levy MD, Abstract Co-Author: Nothing to Disclose

Daniel L. Rubin MD, Presenter: Grant, General Electric Company

Identifying, tracking and reasoning about cancer lesions is a key task for assessing cancer treatment response, but currently this is a laborious, error-prone process. Our goal was to develop methods to enable computerized reasoning about tumor lesions to automate and reduce variation in imaging assessment of treatment response.

We adapted the Annotation and Image Markup (AIM) caBIG information model to encode the semantic information related to cancer imaging findings, and we implemented AIM as a plugin to the Osirix image viewing workstation. We developed a methodology and a suite of tools for transforming AIM image annotations in XML into Web Ontology Language (OWL), and an OWL ontology for computer reasoning with AIM annotations for tumor lesion assessment. We also defined and implemented two computer processing procedures that perform image-based reasoning on OWL instances derived from images: 1) calculation of the length of each lesion, and 2) classification of lesions as measurable and non-measurable based on semantic information about the location, type, and calculated length of each lesion. To evaluate our system, we used our tool to annotate 116 images from 10 cancer patients who had serial imaging studies and to infer response characteristics of cancer lesions.

The inferences produced by our system were reviewed by an oncologist who confirmed that they were correct based on the raw image annotation information. The image annotations could be acquired as part of routine clinical workflow, and automated assessment of lesions to calculate tumor response took a fraction of time as when it was done without too support. In qualitative terms, the oncologist believed our system will streamline image-based evaluation of cancer patients.

Our system enables automated inference of semantic information about cancer lesions in images. This approach could improve the accuracy and reproducibility of image-based cancer treatment response assessment.

Our methods to enable image-based computer reasoning could improve the accuracy and efficiency of assessing cancer patients.

Levy, M, Rubin, D, Semantic Reasoning with Image Annotations for Tumor Assessment.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12029522.html