Automatic Segmentation and Classification of Masses from Digital Mammograms

Breast cancer is one of the leading causes of death among female cancer patients. Mammography is the most efficient method for the early detection of abnormalities that are associated with breast cancer. Masses and micro calcifications are the most popular abnormalities that indicate breast cancer. The proposed paper intends to develop an automated system for assisting the analysis of digital mammograms. First, a preprocessing step is applied to enhance images followed by a segmentation step that is based on morphological operations and Otsu's thresholding techniques. Thereafter, shape features are extracted from the segmented region and used in the classification process. Finally, the classification step to classify the segmented shape as round, oval, lobular, or irregular. The algorithm is tested using 270 mammogram images from the Women Health Care Program (WHC) and 142 publicly available images from the Digital Database for Screening Mammography (DDSM). Results show that the proposed technique effectively detects and segments masses from mammogram images. The shape of segmented masses is classified into either round, oval, lobular, or irregular. Round and oval shapes are classified with 100% accuracy while lobular and irregular shapes results in accuracy of 93% using the ANN for the WHC dataset and 100%. On the other hand, accuracy for images from the DDSM is 100% and 91.3% respectively.

Hali tarjima qilinmagan