Optimal filter-based detection of microcalcifications

This paper deals with the problem of texture feature extraction in digital mammograms. We use the extracted features to discriminate between texture r epresenting clusters of microcalcifications and texture representing normal tissue. Having a two-class problem, we suggest a. texture feature extracti on method based on a single filter optimized with respect to the Fisher cri terion. The advantage of this criterion is that it uses both the feature me an and the feature variance to achieve good feature separation. Image compr ession is desirable to facilitate electronic transmission and storage of di gitized mammograms. In this paper, we also explore the effects of data comp ression on the performance of our proposed detection scheme. The mammograms in our test set were compressed at different ratios using the Joint Photog raphic Experts Group compression method. Results from an experimental study indicate that our scheme is very well suited for detecting clustered micro calcifications in both uncompressed and compressed mammograms. For the unco mpressed mammograms, at a rate of 1.5 false positive clusters/image our met hod reaches a true positive rate of about 95%, which is comparable to the b est results achieved so far. The detection performance for images compresse d by a factor of about four is very similar to the performance for uncompre ssed images.