Jack Sklansky, Presenter: Nothing to Disclose

Chester Ornes MSc, Abstract Co-Author: Nothing to Disclose

Jeffrey Klein MD, Abstract Co-Author: Nothing to Disclose

Gary Levine MD, Abstract Co-Author: Nothing to Disclose

Detecting abnormal growth in sequences of  mammograms of dense breasts, and using this growth to train relational maps, could overcome a major controversy in the management of mammographic screening -- namely the screening of women aged 40-to-49 years.  

We describe two patented inventions to improve the sensitivity and specificity of mammographic screening of dense breasts: FIDUCIAL SYSTEMS FOR MAMMOGRAPHY and  VISUAL NEURAL CLASSIFIER. Using FIDUCIAL SYSTEMS,  invisible radiopaque tattoos implanted in the breast and a digital model of the elastic breast enable repositioning and  matching of the tattoos in sequenced breast images.  Subtracting tattoo-matched breast images reveals abnormal  growth and suppresses most of the rest of the breast tissue. Using VISUAL NEURAL CLASSIFIER, a large database of breast regions of interest (ROIs) is mapped into a two-dimensional "relational map" of labeled dots. Each dot represents one image in the database. The dots are arranged in clusters, such that closely spaced dots denote images that are medically and visually similar to each other. A learning system produces clusters of dots that reflect the user-radiologist's BI-RADS descriptors, such as "pleomorphic" and "segmental."  A second learning system trains the clusters to reflect the detected tissue growth. A third learning system trains the clusters to reflect biopsy assessments.      

We constructed a breast phantom from plastic foam and amicrometric precision positioning stage. We x-ray imaged the phantom in the stage, removed the phantom from the stage, insertedstraight pinsto mimicgrowths, andx-ray imaged the restagedand repositioned phantom. Subtracting the first image from the second demonstratedtheFIDUCIAL SYSTEMS' ability to suppressfiducial marks and background noise and enhance the visibility of new growth.Recent tests on 25 anonymouscalcification cases at Hoag Hospital Newport Beachsupport using relational maps to match queryROIs to similar proven cases.

We constructed a breast phantom from plastic foam and a micrometric precision positioning stage.  We x-ray imaged  the phantom in the stage, removed the phantom from the stage, inserted straight pins to mimic growths, and x-ray  imaged the restaged and repositioned phantom.  Subtracting the first image from the second demonstrated the FIDUCIAL SYSTEMS' ability to suppress fiducial marks and background noise and enhance the visibility of new growth. Recent tests on 25 anonymous calcification cases at Hoag Hospital Newport Beach support  using relational maps to match query ROIs to similar proven cases.

Sklansky, J, Ornes, C, Klein, J, Levine, G, Computer-aided Mammographic Screening of Dense Breasts.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12022132.html