Simon Han, Presenter: Nothing to Disclose

William Hsu PhD, Abstract Co-Author: Nothing to Disclose

Corey W. Arnold, Abstract Co-Author: Nothing to Disclose

Daniel Jason Aaron Margolis MD, Abstract Co-Author: Research Grant, Siemens AG

Alex Anh-Tuan Bui MS, PhD, Abstract Co-Author: Nothing to Disclose

Dieter Roland Enzmann MD, Abstract Co-Author: Nothing to Disclose

Prostate imaging is increasingly being used to detect, stage, and manage individuals with prostate cancer given its ability to characterize lesions in a non-invasive and objective way. However, the sensitivity and specificity of these imaging exams are not well studied. We present a software tool that automatically assesses the concordance between radiology and pathology report findings from clinical text documents. The system was previously demonstrated to work in breast imaging (BIRADS), and we have adapted the tool to work for prostate exams at our institution. While recommended radiological reporting guidelines exist for prostate (PIRADS), this metric is relatively recent and not widely adopted. As such, we extract relevant elements from retrospective imaging reports to determine the radiological interpretations and compare the results to pathology findings in order to generate summarizing visualizations to assess the utility of imaging in diagnosing prostate cancer.

The Radiology Quality Assurance (RadQA) software was implemented for prostate imaging cases at a large medical institution. 114 patients who have undergone targeted prostate biopsies or radical prostatectomies with resulting pathology results have been processed by the tool.

The system automatically generates statistics (e.g., number of biopsied patients, concordance to pathology reports) in the dashboard. Users can specify filters such as time period, modality, and quality. Since prostate imaging reports are semi-structured, a challenge is to extract relevant elements from these reports and correctly match lesions from one report to another.

Our software tool can assist in departmental efforts to improve the quality and value of information from radiology interpretations by objectively assessing its accuracy in comparison with downstream results. Currently, we determine the radiological interpretation by mining text reports, but as structured reporting and scoring systems (e.g., PIRADS) become used more widely, we will improve the ability to make meaningful comparisons. In addition to prostate, we are exploring extending the framework to other domains such as lung cancer.

Han, S, Hsu, W, Arnold, C, Margolis, D, Bui, A, Enzmann, D, RadQA: Automated Quality Control of Radiological Interpretations in Prostate Cancer.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14011309.html