Purpose: A retrospective evaluation of the ability of computer-aided detection (CAD) ability to identify breast carcinoma in standard mammographic projections.
Materials and methods: Forty-five biopsy proven lesions in 44 patients imaged digitally with CAD applied at examination were reviewed. Forty-four screening BIRADS category 1 digital mammography examinations were randomly identified to serve as a comparative normal/control population. Data included patient age; BIRADS breast density; lesion type, size, and visibility; number, type, and location of CAD marks per image; CAD ability to mark lesions; needle core and surgical pathologic correlation.
Results: The CAD lesion/case sensitivity of 87% (n = 39), image sensitivity of 69% (n = 31) for mediolateral oblique view and 78% (n = 35) for the craniocaudal view was found. The average false positive rate in 44 normal screening cases was 2.0 (range 1-8). The 2.0 figure is based on 88 reported false positive CAD marks in 44 normal screening exams: 98% (n = 44) lesions proceeded to excision; initial pathology upgraded at surgical excision from in situ to invasive disease in 24% (n = 9) lesions.
Conclusion: CAD demonstrated potential to detect mammographically visible cancers in standard projections for all lesion types.