Artificial Intelligence Assisted Left Ventricular Diastolic Function Assessment and Grading：Multi-view versus Single-view.

Researchers

Journal

Modalities

Models

Abstract

Clinical assessment and grading of left ventricular diastolic function (LVDF) requires quantification of multiple echocardiographic parameters interpreted according to established guidelines which depends on experienced clinicians and is time-consuming. We aim to develop an artificial intelligence (AI) assisted system to facilitate the clinical assessment of LVDF.We used 1304 studies (33404 images) to develop a view classification model to select 6 specific views required for LVDF assessment. We used 2238 studies (16794 2D images and 2198 Doppler images) to develop 2D and Doppler segmentation models respectively to quantify key metrics of diastolic function. We used 2150 studies with definite LVDF labels determined by 2 experts to train single-view-based classification models by AI interpretation of strain metrics or video. The accuracy and efficiency of these models were tested in an external dataset of 388 prospective studies.The view classification model identified views required for LVDF assessment with a good sensitivity (>0.9) and view segmentation models successfully outlined key regions of these views with Inter Over Union (IoU) >0.8 in internal validation dataset. In external test dataset of 388 cases, AI quantification of 2D and Doppler showed narrow limits of agreement (LOA) compared with the two experts (e.g., LVEF: -12.02% to 9.17%; E/e’: -3.04 to 2.67). These metrics were used to detect LV diastolic dysfunction (DD) and grade DD with a accuracy of 0.9 and 0.92 respectively. Concerning single-view-based method, the overall accuracy of DD detection was 0.83 and 0.75 by strain-based and video-based model; and the accuracy of DD grading was 0.85 and 0.8 respectively. These models could achieve the diagnosis and grading of LVDD in a few seconds, greatly saving time and labor.AI models successfully achieved LVDF assessment and grading that compared favorably to human experts reading according to guideline-based algorithms. Moreover, under conditions when Doppler variables were missing, AI models could provide assessment by interpreting 2D strain metrics or videos from a single view. These models had potential to save labor, cost and facilitate workflow of clinical LVDF assessment.Copyright © 2023. Published by Elsevier Inc.