Identification of Abnormal Masses in Digital Mammogram Using Statistical Decision Making
Indra Kanta Maitra, Samir Kumar Bandyopadhyay · 2017
Diagnostic imaging provides an efficient way for noninvasive mapping of the human anatomy. The increasing volume of data produced by diagnostic imaging can be efficiently managed by computer aided detection/diagnosis (CAD) to assist medical practitioners in image interpretation. CAD is a young interdisciplinary technology combining of artificial intelligence and digital image processing with medical imaging. In recent times, CAD has become a part of clinical diagnosis specially to detect structural abnormalities like tumors. Mammography has been proven to be an accurate and dependable screening methodology for early breast tumors detection. In this chapter, an automated segmentation technique has been proposed for digital mammogram to detect abnormal masses (i.e., tumours). A precise analysis of CAD to find out the existence of abnormalities from mammogram images depends on an accurate segmentation algorithm. The accuracy of an automatic segmentation algorithm requires standardization (i.e., preparation of image and preprocessing of medical images that are mandatory, distinct, and sequential). The detection method is based on a modified seeded region-growing algorithm (SRGA) followed by a step-by-step statistical elimination method. Finally, a decision-making system is proposed to isolate masses in mammogram. The receiver operating characteristic (ROC) analysis suggests that the algorithmic accuracy is 96%, whereas sensitivity and specificity are 97.6% and 88.6%, respectively. The proposed method has been intensively tested with standard data sets that show false-positive (FP) and true-negative (TN) cases under an acceptable threshold.