While CHF can be deemed probable in patients with particular symptoms, risk factors, electrocardiogram findings, and laboratory results, a definitive diagnosis can only be made through echocardiography or a cardiac MRI. Echocardiography requires skilled personnel to administer the test, and then a specialist physician (usually a cardiologist or radiologist) must read the study and visually assess how well the heart is pumping. This is usually done by estimating the ejection fraction (EF), which is the fraction of blood the left ventricle ejects during its contraction. An EF of 65% is considered normal, 40% indicates heart failure, and 10-15% is seen in advanced stages of CHF. The following diagram was taken from an echocardiogram, which shows the four chambers of the heart. You can imagine that it may be unreliable to quantify heart function using the fuzzy images produced by sound waves:

A cardiac MRI, while more expensive, is more accurate in measuring EF and is considered the gold standard for CHF diagnosis; however, it requires a cardiologist to spend up to 20 minutes reading an individual scan. In the following diagram, we see:

• A) An illustration of the heart with the imaging plane used for cardiac MRI also shown.
• B) A 3-D angiogram of the heart (an angiogram is a study in which dye is injected into the bloodstream while images are taken to better visualize the blood vessels).
• C), D), and E) images of normal, damaged, and ischemic left ventricles, respectively, obtained from cardiac MRIs:

Given the nontrivial nature of CHF diagnosis, again we ask the same questions we asked previously: Can new risk factors be discovered and better performance be achieved with respect to CHF diagnosis by using machine learning algorithms?