Computer-Aided Detection of Breast Cancer on Mammograms
Breast cancer is the leading cause of death in women. Masses and microcalcification clusters are important early signs of breast cancer. Precise early detection can effectively mitigate the mortality rate caused by microcalcification in breast. However, it is difficult to identify the cancer from normal breast tissues because of their subtle appearance and unclear skin line. Mammogram-based X-ray study is considered as one of the most reliable and accurate methods to prescreen breast malignancy. However, analysis of the mammogram images by radiologists is manpower consuming, time consuming, and prone to errors. A computer-based machine learning algorithm helps the radiologist in identification of unnatural tissues in an easy way. We look into a new classification idea for detection of microcalcification in digital breast images using extreme learning machine. This classification method is done to detect and differentiate microcalcifications from the normal tissue using textural features and is compared with different feature vectors extracted using gray level spatial dependence matrix (GLSDM), SURF filter, and Gabor filter based techniques. This algorithm extracts features from images using extracting laws texture energy measures and classifying the dubious regions by learning a pattern classifier. This classification method is experimented on MIAS database. The classifier is learned with the above mentioned features and the results denote that extreme learning machine with SURF produces better classification prediction accuracy (95%) with an evidential reduction in learning time than the other artificial neural methods like naïve Bayes and support vector machine.