Clinical phenotypes among patients with normal cardiac perfusion using unsupervised learning: a retrospective observational study

Abstract

Background: Myocardial perfusion imaging (MPI) is one of the most common cardiac scans and is used for diagnosis of coronary artery disease and assessment of cardiovascular risk. However, the large majority of MPI patients have normal results. We evaluated whether unsupervised machine learning could identify unique phenotypes among patients with normal scans and whether those phenotypes were associated with risk of death or myocardial infarction.

Methods: Patients from a large international multicenter MPI registry (10 sites) with normal perfusion by expert visual interpretation were included in this cohort analysis. The training population included 9849 patients, and external testing population 12,528 patients. Unsupervised cluster analysis was performed, with separate training and external testing cohorts, to identify clusters, with four distinct phenotypes. We evaluated the clinical and imaging features of clusters and their associations with death or myocardial infarction.

Findings: Patients in Clusters 1 and 2 almost exclusively underwent exercise stress, while patients in Clusters 3 and 4 mostly required pharmacologic stress. In external testing, the risk for Cluster 4 patients (20.2% of population, unadjusted hazard ratio [HR] 6.17, 95% confidence interval [CI] 4.64-8.20) was higher than the risk associated with pharmacologic stress (HR 3.03, 95% CI 2.53-3.63), or previous myocardial infarction (HR 1.82, 95% CI 1.40-2.36).

Interpretation: Unsupervised learning identified four distinct phenotypes of patients with normal perfusion scans, with a significant proportion of patients at very high risk of myocardial infarction or death. Our results suggest a potential role for patient phenotyping to improve risk stratification of patients with normal imaging results.

Funding: This work was supported by the National Heart, Lung, and Blood Institute at the National Institutes of Health [R35HL161195 to PS]. The REFINE SPECT database was supported by the National Heart, Lung, and Blood Institute at the National Institutes of Health [R01HL089765 to PS]. MCW was supported by the British Heart Foundation [FS/ICRF/20/26002].

Keywords: Cluster analysis; Coronary artery disease; Machine learning; Myocardial perfusion.

Fig. 1

Population flow diagram outlining patient inclusions and exclusions. Visual normal perfusion was defined as summed stress score of zero or expert reader interpretation of normal if summed stress score was not available. Refer to Supplemental Table S1 for a per-site exclusions summary.

Fig. 2

Overview of study design and findings. Unsupervised machine learning was used to identify four unique clusters of patients with normal myocardial perfusion. Patients in the highest risk cluster had 6-fold higher risk of death or myocardial infarction in the training and external testing populations.

Fig. 3

Training and external testing populations projected into the reduced embedding space. Components of the embedding space are independent summary measures that combine multiple input parameters, which are determined by the non-linear dimensionality reduction process. There is a clear separation of groups in the training population with Silhouette score of 0.881. In the external testing population, there is still separation between groups, but with a few outliers, the Silhouette score was 0.852.

Fig. 4

Shapley additive explanation (SHAP) values for the top 24 features and all remaining features were generated using the Cluster-Shapley method. ECG—electrocardiogram, MPHR—maximum predicted heart rate.

Fig. 5

Kaplan–Meier curves for death or myocardial infarction in the training population. The radial plots visualize differences between clusters compared to the entire training population (inside/outside of orange circle). A higher proportion of patients, or larger values, is identified as an over-represented trait (outside). The opposite is true for an under-represented trait (inside). Hazard ratio (HR) and 95% confidence intervals are shown for each Cluster compared to Cluster 1. LVEDV—left ventricular end-diastolic volume, LVEF—left ventricular ejection fraction, TPD—total perfusion deficit.

Fig. 6

Kaplan–Meier curves for death or myocardial infarction in the external testing population. The radial plots visualize differences between clusters compared to the entire external testing population (inside/outside of orange circle). A higher proportion of patients, or larger values, is identified as an over-represented trait (outside). The opposite is true for an under-represented trait (inside). Hazard ratio (HR) and 95% confidence intervals are shown for each Cluster compared to Cluster 1. LVEDV—left ventricular end-diastolic volume, LVEF—left ventricular ejection fraction, TPD—total perfusion deficit.