AI Model Shows Reduced Blood Flow Leads to Adverse Cardiovascular Events

James Moon, MD

Using an artificial intelligence (AI) model, investigators were able to determine that patients with reduced blood flow were more likely to have adverse health outcomes, including death, heart attack, stroke, and heart failure, according to recent findings of a multicenter study.

The technology could help predict chances of death, heart attack, and stroke, and providers can use the information to make better informed treatment recommendations to improve a patient’s blood flow.

James Moon, MD, and a team of investigators studied patients with both suspected and known coronary artery disease who were referred for perfusion assessment. The team used an AI approach that derived global and regional stress and myocardial blood flow and perfusion reserve.

Moon, from the Institute of Cardiovascular Science at University College London, and colleagues took routine cardiovascular magnetic resonance scans from 1049 patients at St. Bartholomew’s Hospital and the Royal Free Hospital. The cardiovascular magnetic resonance scans were analyzed by an accredited cardiologist using commercially available software. The cardiologist recorded left ventricle systolic, diastolic volume, ejection fraction, and the presence and distribution of late gadolinium enhancement.

Perfusion maps were generated automatically during the scans.

The AI performed automatic segmentation of the left ventricle cavity and myocardium. Using a convolutional neural network, the technology delineated the left ventricle cavity and myocardium excluding myocardial fat and papillary muscles.

Global myocardial blood flow was then automatically calculated as an average of all pixels and global myocardial perfusion rate as the ratio of stress to rest myocardial blood flow.

The investigators performed Cox proportional hazard regression analysis to determine whether perfusion data—stress myocardial blood flow and perfusion rate—were associated with death and major adverse cardiovascular events.

Among the 1049 patients included in the analysis, 1018 had myocardial perfusion rate data. The mean age of the patients was 60.9+13 years and 702 (67%) were male. Nearly 300 (28%) had diabetes mellitus; 630 (60%) had hypertension; 510 (49%) had dyslipidemia; 318 (30%) had previous revascularization; 360 (34%) had a smoking history; 63 (6%) had a previous stroke or transient ischemic attack; 141 (13%) had atrial fibrillation (AF); and 108 (10%) had a current or previous history of cancer.

Mean ejection fraction was 60%+13%, 309 (30%) patients had infarct pattern and 113 (13%) had non-infarct pattern late gadolinium enhancement. Mean stress myocardial blood flow was 2.06+.71 ml/g/min and perfusion rate was 2.48+.82.

Overall, there were 42 (4%) deaths during the following period of 605 days (interquartile range 464-814 days) and 188 major adverse cardiovascular events in 174 (16.6%) patients. Myocardial blood flow was lower in patients who died (1.70+.65 vs 2.08+0.71ml/g/min, P = .001), along with perfusion rate (1.97+.74 vs 2.5+0.81, P <.001).

Major events included 28 (2.7%) myocardial infarctions, 10 (.95%) strokes, 18 (1.7%) heart failure admissions, and 127 (12.1%) late revascularizations. Patients that had such an event were more commonly older males with more often prior revascularizations and were more likely to have diabetes, hypertension, dyslipidemia, a previous stroke or transient ischemic attack, and a positive smoking history.

For each 1 ml/g/min decrease in stress myocardial blood flow, the adjusted hazard ratio for death was 1.93 (95% CI, 1.08-3.48; P = .028) and 2.14 (95% CI, 1.58-2.9; P <.0001) for major adverse cardiovascular events. Each 1 unit decrease in myocardial perfusion rate was associated with an adjusted hazard ratio of 2.45 (95% CI, 1.42-4.24; P = .001) for death and 1.74 (95% CI, 1.36-2.22; P <.0001) for major adverse cardiovascular events.

The AI was able to predict which patient might die or suffer events better than a doctor could with traditional approaches.

“This study demonstrates the growing potential of artificial intelligence-assisted imaging technology to improve the detection of heart disease and may move clinicians closer to a precision medicine approach to optimize patient care,” Peter Kellman, PhD, from the National Heart, Lung, and Blood Institute at the National Institutes of Health, said in a statement.

The study, “The Prognostic Significance of Quantitative Myocardial Perfusion: An Artificial Intelligence Based Approach Using Perfusion Mapping,” was published online in the journal Circulation.