Enhanced diagnostic accuracy of mammograms on a mobile device
Abstract
With the death of a woman every 13 minutes in the US, and one every minute worldwide, due to breast cancer, the need for early detection cannot be overstated. Mammography is a boon for both early detection and screening of breast tumors. It is an imaging system that uses low dose (9mrem) x-rays for examining the breasts, by the electrons reflected from the tissues (thermoelectric effect). However, there are 20% false positives and 10% false negatives in current practice. Hence, there is a critical need for enhancing the accuracy of these mammograms. Towards this, this thesis was aimed at enhancing the current diagnostic accuracy of digital mammograms using the industry standard simulation software tool, MATLAB. For this purpose, the publicly available dataset MIAS was used. Image processing techniques, such as wavelet, statistical and feature analysis and pattern recognition algorithms, such as Bayes', Naive Bayes', k-nearest neighbor (kNN), Support Vector Machines (SVM), and Artificial Neural Network (ANN) were utilized to enhance the diagnostic accuracy. The results indicated up to 95% accuracy, compared to 70% at present. The proposed solution has proven to be an effective way of detecting breast cancer early in different types of breast tissues. The entire solution is implemented as a smartphone application to serve as a second opinion for clinical use.
Degree
M.S.
Advisors
Sundararajan, Purdue University.
Subject Area
Computer Engineering|Biomedical engineering|Electrical engineering|Medical imaging
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.