Pragya Ahuja Dang MBBS, Presenter: Nothing to Disclose

Mannudeep Karanvir Singh Kalra MD, Abstract Co-Author: Research grant, General Electric Company

Michael Austin Blake MBBCh, Abstract Co-Author: Nothing to Disclose

Raul Nirmal Uppot MD, Abstract Co-Author: Nothing to Disclose

Thomas Schultz, Abstract Co-Author: Nothing to Disclose

Keith J. Dreyer DO, PhD, Abstract Co-Author: Employee, Perceptics, LLC Medical Advisor, Perceptics, LLC Medical Advisor, Agfa-Gevaert Group Medical Advisor, FUJIFILM Holdings Corporation Medical Advisor, General Electric Company Medical Advisor, McKesson Corporation Medical Advisor, AuntMinnie.com Medical Advisor, AMICAS, Inc Medical Advisor, Dynamic Imaging, LLC Medical Advisor, Ascom Holding AG Medical Advisor, Bracco Group Medical Advisor, Merge Healthcare Medical Advisor, Emageon, Inc Medical Advisor, RCG HealthCare Consulting Medical Advisor, Valley Radiology Medical Associates, Inc Medical Advisor, The Elizabeth Wende Breast Clinic Medical Advisor, ISCI Medical Advisor, Siemens AG Medical Advisor, Barco nv Medical Advisor, Hue AS Medical Advisor, Planar Systems, Inc Medical Advisor, Vital Images, Inc Medical Advisor, Commissure, Inc Medical Advisor, TeraRecon, Inc Medical Advisor, Mercury Computer Systems, Inc Medical Advisor, IBM Corporation Medical Advisor, Hewlett-Packard Company Medical Advisor, EMC Corp Medical Advisor, Phase Forward Incorporated Medical Advisor, Winchester Systems, Inc Medical Advisor, Dell Inc Medical Advisor, Eastman Kodak Company Medical Advisor, Amirsys, Inc Medical Advisor, Reed Elsevier Committee member, Diagnostic Imaging Committee member, AuntMinnie.com Committee member, Imaging Economics Author, Springer Science+Business Media Deutschland GmbH Shareholder, Microsoft Corporation Shareholder, Intel Corporation Shareholder, IBM Corporation Shareholder, Hewlett-Packard Company Shareholder, Dell Inc Shareholder, General Electric Company Shareholder, Siemens AG Shareholder, Google Inc

A current Natural Language Processing (NLP) program can accurately classify radiology reports on the basis of presence or absence of findings but does not take into account the clinical information of the patient. We determined the clinical significance of radiology findings classified as positive (F+) or negative (F-) in abdominal CT reports in a large radiology report database by two different radiologists.

We selected 150 consecutive abdomen CT reports (n=150 patients, M:F=76:74, age range=2-92) categorized as F+ by NLP and 50 reports (M:F=26:24, age range=4-88) categorized as F- by NLP irrespective of clinical indications and representing the proportion of outpatients and inpatients in our practice (5:1). These reports were randomized and presented to 2 radiologists, not aware of NLP classification. Each radiologist graded the reports for clinical significance on a 5-point scale based on clinical indication and electronic medical records. Clinically significant findings (C+) were further classified into 4 finding subcategories (1=none, 2=stable, 3=change from prior, 4=new). The proportions of C+ reports in those classified as F+ and F- were determined. The F+ and F- as classified by NLP in abdominal CT reports in the entire database (n=279,689, 1996-2007) were compared with the proportion of reports with and without C+.

There was excellent agreement between the radiologists (171/200, 85.5%). Of the reports classified as F+, radiologists reported that 34 (22.7%) reports had probable C+ and 107 (71.3%) had definite C+. Of 50 F- reports, 4% reports had no, probably no or indeterminate C+. Of the probably or definitely C+ reports, most (93.2 %) were attributable to the clinical indication and about 2/3rd were new or change from prior findings. Overall, 5% reports were graded as not clinically significant by radiologists whereas 22% reports were classified by NLP as no finding or no change from prior.

Contrary to current NLP rules, most CT reports with no findings or no change from prior are categorized as clinically significant by radiologists when incorporating the patient’s clinical information. However, there is a good correlation between positive radiology findings by NLP and clinically significant radiology reports.

Regardless of differences between radiologists and NLP, the yield for clinically significant findings from abdominal CT reports ranges from 80-95%.

Dang, P, Kalra, M, Blake, M, Uppot, R, Schultz, T, Dreyer, K, Clinical Relevance of Radiology Findings Extracted with NLP in Abdominal CT.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6011005.html