Standard image processing offers optimal calcification detection in digital mammography

Image processing has a significant effect on the detection of calcification clusters in digital mammography, with standard image processing outperforming low-contrast and film-screen image-processing algorithms, according to a study published in the August issue of the American Journal of Roentgenology.

While there was no significant difference in the number of false-positives for noncalcification cancers based on image processing, there was a significant decrease in the number of false-positives for calcification clusters using standard processing, according to Lucy M. Warren, MD, of the National Coordinating Centre for the Physics of Mammography, Royal Surrey County Hospital in Guildford, U.K., and colleagues.

“The change in the number of false-positives in these images indicates that benign features that did not appear suspicious enough to be recalled in the clinic can appear suspicious enough to be recalled when a different type of image processing has been applied,” wrote the authors.

Warren and colleagues noted that previous studies of mammography image processing had a number of limitations in their methods, and they set about comparing processing algorithms using an objective measure: jackknife-alternative free-response receiver operating characteristic (JAFROC) analysis.

A total of 270 pairs of breast images were collected for the retrospective observer study, all from systems using the same detector brand. Of these images, 80 image pairs contained subtle malignant masses, 30 contained biopsy-proven benign lesions, 80 featured simulated calcification clusters and 80 were normal scans with no cancer.

The image pairs were processed with a standard algorithm featuring full enhancement, a low contrast algorithm with intermediate enhancement, and a pseudo-film-screen with no enhancement. Results showed the JAFROC figure of merit decreased from 0.65 with standard processing to 0.63 and 0.61 with low-contrast and film-screen processing, respectively.

While the type of processing did have a significant effect on calcification detection, the difference was smaller than the change in detection based on factors such as detector type or dose, according to Warren and colleagues.