Mammography Study Example of Management Accuracy

X-ray-mammography is the most sensitive technique for detecting breast cancer [2], with a reported sensitivity of 8595% for detecting small lesions. Most noninvasive ductal carcinomas, or DCIS, are characterized by tiny nonpalpable calcifications detected at screening mammography [9,16,29]. Traditional mammography is essentially analog photography using X-rays in place of light and analog film for display. Mammography machines based on direct digital acquisition exist, and the review is in process of FDA approval for market. The study discussed here, however, employed only digitized analog films. The studies were digitized using a Lumisys Lumiscan 150 at 12bpp with a spot size of 50 microns. After compression, the images were put on hardcopy film. The films were printed using a Kodak 2180 X-ray film printer, a 79-micron, 12-bit gray-scale printer that writes with a laser diode of 680-nm bandwidth.

Images were viewed on hardcopy film on an alternator by judges in a manner that simulates ordinary screening and diagnostic practice as closely as possible, although patient histories and other image modalities were not provided. Two views were provided of each breast (CC and MLO), so four views were seen simultaneously for each patient. Each of the judges viewed all the images in an appropriately randomized order over the course of nine sessions. Two sessions were held every other week, with a week off in between. A clear overlay was provided for the judge to mark on the image without leaving a visible trace. For each image, the judge either indicated that the image was normal, or, if something was detected, had an assistant fill out the Observer Form using the American College of Radiology (ACR) Standardized Lexicon by circling the appropriate answers or filling in blanks as directed. The Observer Form is given in Figs 11-13 below. The instructions for assistants and radiologists along with suggestions for prompting and a CGI web data entry form may be found at the project Web site http://www-isl.stanford.edu/ ~gray/armyhtml. The judges used a grease pencil to circle the detected item. The instructions to the judges specified that ellipses drawn around clusters should include all microcalcifications seen, as if making a recommendation for surgery, and outlines drawn around masses should include the main tumor as if grading for clinical staging, without including the spicules (if any) that extend outward from the mass. This corresponds

ID number_

Reader initials

Session number

Case number

Mammograms were of ( Left Right Both ) breast(s). Subjective rating for diagnostic quality:

Left CC

Left MLO

Right CC

Right MLO

0 0

Post a comment