Abstract

Breast cancer is an essential health issue and more than one million women die of breast cancer each year in the world. Primary prevention seems impossible since the causes of this disease still remain unknown. Early detection and diagnosis is the key for breast cancer control. Clusters of microcalcifications in mammogram have been mainly targeted as an early sign of breast cancer and their earliest detection is vital to reduce the mortality rate. Since the size of microcalcification is very tiny and may be overlooked by the observing radiologist, a Computer Aided Diagnosis (CAD) system has been developed for efficient microcalcification detection and it eliminates the operator dependency. In order to determine the presence of microcalcification clusters in the mammogram, special attention is paid to the analysis of the structure and brightness of the mammogram tissues. The detection of microcalcification clusters is achieved by the following computerized approach. The nodular structured microcalcifications in the abnormal mammogram image are detected based on multiscale products of eigenvalues of the Hessian matrix. The detected image contains calcifications along with background information. To eliminate the unnecessary background information, the response image coming out from Hessian matrix approach is passed to the thresholding technique such as probability density function based Tsallis entropy, in which the potential microcalcifications are segmented efficiently. The proposed method is evaluated with the DDSM, MIAS and UCSF databases, which included a wide spectrum of difficult-to-detect cases. The detection performance of the proposed method has been evaluated by using 234 mammograms containing 171 microcalcification clusters. The detection method has a true positive ratio of 97.08% with 0.45 false positives per image. The proposed algorithm is positively evaluated through the clinical study. The main merit of the system is that it maintains the shape and size of microcalcifications.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call