Automated heart abnormality detection using sparse linear classifiers.

C ardiovascular disease (CVD) is a global epidemic that is the leading cause of death worldwide (17 million deaths per year) [8]. It is the single largest contributor to " disability adjusted life years " (DALYs): 10% of DALYs in low-and middle-income nations and 18% of DALYs in high-income nations. Hence, the World Health Organization and the Centers for Disease Control agree that CVD is no longer an epidemic but a pandemic. In the United States, CVD accounted for 38% of all deaths in 2002 [7] and was the primary or contributing cause in 60% of all deaths. Coronary heart disease (CHD) accounts for more than half of CVD deaths (roughly 7.2 million deaths worldwide every year, and one of every five deaths in the United States), and it is the single largest killer in the world. It is well known that early detection (along with prevention) is an excellent way of controlling CHD. CHD can be detected by measuring and scoring the regional and global motion of the left ventricle (LV) of the heart. It typically results in wall-motion abnormalities [i.e., local segments of the LV wall move abnormally (move weakly, not at all, or out of sync with the rest of the heart)], and sometimes motion in multiple regions, or indeed the entire heart, is compromised. The LV can be imaged in a number of ways. The most common method is the echocardiogram, which is an ultrasound video of different two-dimensional cross sections of the LV. Unfortunately, echocardiograms are notoriously difficult to interpret, even for the best of physicians. Inter-observer studies have shown that even world-class experts agree on their diagnosis only 80% of the time [12], and intra-observer studies have shown a similar variation when the expert reads the same case twice at widely different points in time. There is a tremendous need for an automated " second-reader " system that can provide objective diagnostic assistance, particularly to the less-experienced cardiologist. In this article, we address the task of building a computer-aided diagnosis system that can automatically detect wall-motion abnormalities from echocardiograms. We provide some medical background on cardiac ultrasound and the standard methodology used by cardiologists to score wall-motion abnormalities. We also describe our real-life dataset, which consists of echocardiograms used by cardiologists at St. Francis Heart Hospital to diagnose wall-motion abnormalities. We then provide an overview of our proposed system, which was built on top …