A machine learning algorithm derived from thoracic phase signals can identify obstructive coronary artery disease (CAD) with the same accuracy as existing functional tests, according to a study published Aug. 8 in PLOS One. The signals can be collected in about three minutes and don’t require the patient to exercise or be exposed to radiation, contrast media or pharmacological stress.
“In the near future, new diagnostic tools based on machine learning will become widely available,” wrote lead author Thomas D. Stuckey, MD, with Cone Health Heart and Vascular Center in Greensboro, North Carolina, and colleagues. “Our analysis of Cardiac Phase Space Tomography suggests that machine-learned algorithms may offer a valuable new method for the assessment of patients with coronary artery disease.”
The researchers developed the algorithm to identify the presence of obstructive CAD, defined by stenosis of 70 percent or greater upon angiography or fractional flow reserve below 0.80. A total of 512 patients who presented to the hospital with chest pain and were referred by a physician for coronary angiography were used to develop the model.
The model was then tested in an additional 94 subjects and showed a sensitivity of 92 percent and specificity of 62 percent upon blind testing. The negative predictive value was 96 percent.
“cPSTA (cardiac space phase tomography analysis) exhibits comparable diagnostic performance to existing functional and anatomical modalities without the requirement of cardiac stress (exercise or pharmacological) and without exposure of the patient to radioactivity,” Stuckey et al. wrote. “This technology may provide a new and efficient technique for assessing the presence of obstructive coronary lesions in patients presenting with chest pain suspected to be of cardiac etiology.”
For the test, seven sensors are placed on the patient’s chest and back while a handheld device collects resting phase signals. Signal data is collected in approximately three minutes, after which about 10 million data points are sent to the cloud and evaluated by an analytic engine using the machine learning algorithm. Results are then shown as a phase space tomography model via a web portal, the researchers said.
Stuckey and colleagues believe this method has the potential to result in faster, safer and less expensive diagnosis of CAD. They pointed out there are more than seven million myocardial perfusion imaging studies performed annually in the U.S. using single photon emission CT (SPECT), with each test costing about $1,000.
That technique normally requires the injection of a radionuclide tracer, the patient to exercise on a treadmill and substantially more time than the phase signal test. In addition, SPECT and CT angiography only identified obstructive CAD in 10 to 12 percent of patients deemed at intermediate risk pretest, according to another study cited by Stuckey et al.
“Little has changed with regard to the accuracy of these technologies in the last decade, and better tools for screening CAD are needed,” they wrote.
Daniel joined TriMed’s Chicago editorial team in 2017 as a Cardiovascular Business writer. He previously worked as a writer for daily newspapers in North Dakota and Indiana.