AI based novel insights into cardiac function utilizing high frame rate echocardiographic scans

Real-time evaluation of myocardial function is critical in various life-threatening conditions encountered in the intensive care units. We have developed a novel echocardiographic method based on the analysis of images at high frame-rate (>120 fps) and on artificial intelligence-assisted short-axis segmentation. The method was validated in sheep (n=4) that underwent successive myocardial infarctions and developed systolic heart failure. An artificial-intelligence model based on U-Net was trained and validated using 850 manually segmented images. The AI-model demonstrated accurate segmentation results with Dice-coefficient of 96.7 +/- 0.69%. The development of systolic dysfunction was associated with a significant prolongation of the isovolumic contraction phase, from 59.4 +/- 2.1ms at baseline to 74 +/- 2.8ms after myocardial infarction. Our study paved the way for the development of fully automated and user-independent extraction of cardiac indices from echocardiography.