Computer-based reading of full-field digital mammograms

Mammograms of 50 patients performed on a full-field digital mammography system were evaluated. The images were judged as hardcopy prints and on 2k x 2.5k high resolution monitors. 19 of the 50 patients had histologically proven carcinomas, the remainder showed no malignancies after a follow-up of 2 years. The mammograms were evaluated with respect to image quality, presence of masses and microcalcifications. All detected lesions were categorized according to the BI-RADS(TM) classification. Time was recorded. Postprocessing in softcopy evaluation (windowing and levelling, zooming, inversion) and additional views like spot views in both methods were also timed. Possible changes in the BI-RADS(TM) classification were included in the evaluation protocol. Diagnostic accuracy of the classification of the malignant lesions according to the BI-RADS(TM) categories showed no significant differences between soft- and hardcopy. More time was needed for the evaluation of the softcopies than for the hardcopies. Additional views were recommended in hardcopy more often than in softcopy evaluation (26 versus 21 for reader 1, 20 versus 18 for reader 2). Postprocessing was used in 9 cases by reader 1 and in all cases by reader 2 resulting in an further increase of the evaluation time. Furthermore softcopy reading enables radiographers to use the advantages of a digital system (e.g. CAD, teleradiology, PACS).