This abstract is published in the Journal of the American Society of Echocardiography (volume 33, number 6, year 2020), as part of the proceedings of the American Society Echocardiography Conference 2020.

Title: Quantification of Left Ventricular Regional Wall Motion: Novel Imaging Features to Predict Coronary Artery Disease

Authors: Hawkes, William and Beqiri, Arian and Mumith, Angela and Parker, Andrew and Upton, Ross and McCourt, Annabelle and Woodward, William and Dockerill, Cameron and Heitner, Stephen and Yadava, Mrinal and others

BACKGROUND: Visual assessment of left ventricular regional wall motion during stress echocardiography is a core tool in identifying myocardial ischaemia in cases of suspected coronary artery disease (CAD). However, this technique suffers from poor inter-operator variability and sub-optimal diagnostic performance. Thus, we sought to develop quantitative features of regional wall motion from echocardiographic images and test their ability to stratify CAD outcomes in patients.

METHODS: 6,742 features were extracted from echocardiographic contours of the left ventricle at end systole and end diastole. These features represent regional geometric and kinematic properties of left ventricular endocardial wall motion. 31 features were identified as potential CAD biomarkers in a cohort of 578 patients from a large prospective multisite trial. ROC curves were then generated to test the ability of these features to predict CAD outcome. To test if grouping features together provided a more robust measure, features were combined into groups of 2 or 3 and patients were re-classified. To evaluate the validity and repeatability of the features to predict CAD outcomes, the classification was replicated on an independent cohort of 154 patients.

RESULTS: Individual features were strong predictors of CAD outcome, with AUROCs ranging from 0.76 to 0.86. Peak sensitivity and specificity were 84% and 79%, respectively. Combining features into groups of two or three resulted in additional improvements in CAD prediction (sensitivity=79%, specificity=86%). Validation of the features on an independent dataset demonstrated excellent reproducibility, with AUROCs of 0.48 to 0.88 (maximum sensitivity=84.4% and specificity=76.1%). Combining features into groups of two or three resulted in incremental improvements in sensitivity and specificity, achieving maximum values of 87% and 85%, respectively.

CONCLUSIONS: Quantitative features of regional wall motion can successfully predict CAD outcome with high precision. These features present and exciting opportunity to develop novel imaging biomarkers of myocardial ischaemia. Furthermore, the numerical nature of these features provide an excellent opportunity for integration with artificial intelligence assisted CAD prediction models, with the potential to provide near real-time data to clinicians during stress echocardiography.

Share on

Share on facebook
Share on twitter
Share on linkedin

Further reading

Ultromics and Mayo Clinic to Use AI to Improve Diagnosis of Heart Failure

●  Heart failure is the leading cause of hospitalization for people over the age of 65 – but diagnosis can be challenging and is often delayed through lack of …

AI Solution that Predicts Heart Disease To Be Implemented In NHS Hospitals

Ultromics’ EchoGo system selected among first winners of new £140m NHS AI in Health and Care Award  A pioneering AI platform that can predict heart disease will be implemented in NHS hospitals as …

What’s to Gain from Strain Imaging

Strain imaging techniques have shown great promise in echocardiography to support early signs of cardiac disease and is considered a more accurate and reproducible quantitative analysis of ventricular function and …