Abstract Mammographic Computer-Aided Detection (CAD) systems provide radiologists with an important second opinion and are now being used clinically. The sensitivity and specificity of these devices continue to improve; however, the variability of false positives (FP) makes them difficult to eliminate with any single global approach. This paper describes a technique for targeting a particular class of FP vascular microcalcifications. Vascular microcalcifications are benign microcalcifications that collect inside blood vessels and may be marked by a CAD system. Radiologists consider such marks to be obvious FP and find them distracting. Although vascular microcalcifications represent a small fraction of FP overall, marking them tends to cause radiologists to lose confidence in the CAD's specificity. For this reason, we found this class of FP to be sufficiently important and unique to develop a filter that eliminates them without sacrificing sensitivity. The filter exploits the fact that vascular microcalcification clusters tend to appear as large curvilinear structures, while true-positive clusters do not have this characteristic.
[1]
F. Winsberg,et al.
Detection of Radiographic Abnormalities in Mammograms by Means of Optical Scanning and Computer Analysis
,
1967
.
[2]
Joachim Dengler,et al.
Segmentation of microcalcifications in mammograms
,
1991,
IEEE Trans. Medical Imaging.
[3]
M. Moskowitz,et al.
Evaluation of mammographic calcifications using a computer program.
,
1975,
Radiology.
[4]
K Doi,et al.
Computer-aided detection of clustered microcalcifications: an improved method for grouping detected signals.
,
1993,
Medical physics.
[5]
Heinz-Otto Peitgen,et al.
Scale-space signatures for the detection of clustered microcalcifications in digital mammograms
,
1999,
IEEE Transactions on Medical Imaging.
[6]
Jeffrey W. Hoffmeister,et al.
Using neural networks to select wavelet features for breast cancer diagnosis
,
1996
.