Abstract Archives of the RSNA, 2011
LL-INS-TU4B
Automatic Extraction of Breast Density Information from Mammography Reports
Scientific Informal (Poster) Presentations
Presented on November 29, 2011
Presented as part of LL-INS-TU: Informatics
Bethany Percha, Abstract Co-Author: Nothing to Disclose
Houssam Nassif, Abstract Co-Author: Nothing to Disclose
Jafi Alyssa Lipson MD, Abstract Co-Author: Nothing to Disclose
Elizabeth S. Burnside MD, MPH, Abstract Co-Author: Nothing to Disclose
Daniel L. Rubin MD, Presenter: Grant, General Electric Company
Breast tissue composition is an important risk factor for susceptibility to breast cancer, and information on breast density could play an important role in the development of classification systems for the early and accurate diagnosis of malignancy. Our goal was to develop an automated method to detect and extract the breast density assessments from free-text mammography reports.
We studied many different mammography reports to understand the variety of ways breast density is reported. We developed an algorithm, based on pattern matching and regular expressions in free text, to automatically detect and extract BI-RADS breast density classes. Using radiologists’ unstructured descriptions of mammograms as its input, the algorithm classifies each report into one of four density classes: fatty (1), fibroglandular (2), heterogeneously dense (3), or extremely dense (4). We evaluated the algorithm’s performance on a total of 600 reports from two different institutions. Two different radiologists reviewed the reports to establish the gold standard.
The algorithm achieved 100% and 99% classification accuracy on the reports from the two institutions, respectively.
Our automated method to extract breast density composition from unstructured mammography reports has high accuracy in our testing, and appears promising. This method could enable epidemiological research by facilitating data-mining of large-scale data sets to correlate breast density with other covariates.
Our method for extracting breast density from reports could enable large-scale coding of mammography reports to enable epidemiological research and decision support based on breast density.
Percha, B,
Nassif, H,
Lipson, J,
Burnside, E,
Rubin, D,
Automatic Extraction of Breast Density Information from Mammography Reports. Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL.
http://archive.rsna.org/2011/11008220.html