M.S. in Engineering
Electrical & Computer Engineering
Henry M. Rowan College of Engineering
Fox Chase Cancer Center
Mammographic radiodensity is one of the strongest risk factors for developing breast cancer and there exists an urgent need to develop automated methods for predicting this marker. Previous attempts for automatically identifying and quantifying radiodense tissue in digitized mammograms have fallen short of the ideal. Many algorithms require significant heuristic parameters to be evaluated and set for predicting radiodensity. Many others have not demonstrated the efficacy of their techniques with a sufficient large and diverse patient database. This thesis has attempted to address both of these drawbacks in previous work. Novel automated digital image processing algorithms are proposed that have demonstrated the ability to rapidly sift through digitized mammogram databases for accurately estimating radiodensity. A judicious combination of point-processing, statistical, neural and contrast enhancement techniques have been employed for addressing this formidable problem. The algorithms have been developed and exercised using over 700 mammograms obtained from multiple age and ethnic groups and digitized using more than one type of X-ray digitizer. The automated algorithms developed in this thesis have been validated by comparing the estimation results using 40 of these mammograms with those predicted by a previously established manual segmentation technique. The automated algorithms developed in this thesis show considerable promise to be extremely useful in epidemiological studies when correlating other behavioral and genetic risk factors with mammographic radiodensity.
Kim, Min Taek, "Automated evaluation of radiodensities in a digitized mammogram database using local contrast estimation" (2004). Theses and Dissertations. 1173.