Computer-aided diagnosis scheme using a filter bank for detection of microcalcification clusters in mammograms

• Published in: IEEE transactions on biomedical engineering 2006-02, Vol.53 (2), p.273-283
• Main Authors: Nakayama, R., Uchiyama, Y., Yamamoto, K., Watanabe, R., Namba, K.
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial Intelligence; Biomedical imaging; Breast cancer; Breast Diseases - diagnostic imaging; Breast Neoplasms - diagnostic imaging; Calcinosis - diagnostic imaging; Cancer detection; Cluster Analysis; Computer aided diagnosis; Discrete wavelet transforms; Discriminant Analysis; Female; Filter bank; Hessian matrix; Humans; Image analysis; Mammography - methods; Matrix decomposition; Medical diagnostic imaging; microcalcification cluster; multiresolution analysis; Pattern Recognition, Automated - methods; Precancerous Conditions - diagnostic imaging; Radiographic Image Interpretation, Computer-Assisted; Radiology; Reproducibility of Results; Retrospective Studies; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Studies
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2005.862536

Mammography is considered the most effective method for early detection of breast cancers. However, it is difficult for radiologists to detect microcalcification clusters. Therefore, we have developed a computerized scheme for detecting early-stage microcalcification clusters in mammograms. We first developed a novel filter bank based on the concept of the Hessian matrix for classifying nodular structures and linear structures. The mammogram images were decomposed into several subimages for second difference at scales from 1 to 4 by this filter bank. The subimages for the nodular component (NC) and the subimages for the nodular and linear component (NLC) were then obtained from analysis of the Hessian matrix. Many regions of interest (ROIs) were selected from the mammogram image. In each ROI, eight features were determined from the subimages for NC at scales from 1 to 4 and the subimages for NLC at scales from 1 to 4. The Bayes discriminant function was employed for distinguishing among abnormal ROIs with a microcalcification cluster and two different types of normal ROIs without a microcalcification cluster. We evaluated the detection performance by using 600 mammograms. Our computerized scheme was shown to have the potential to detect microcalcification clusters with a clinically acceptable sensitivity and low false positives.