Year: 2023
Type: conference-paper
Abstract: The 17-segment model for the left ventricle defined by the American Heart Association is widely used in clinical practice for dividing the geometry of the heart into anatomical parts and for further diagnosis of heart disease.However, the construction of this model from clinical data can require a significant amount of manual work.This study aimed to develop an automatic algorithm for building the 17-segment AHA model on the ventricular meshes.As initial data, our algorithm starts with the surface mesh of the left and right ventricles (LV and RV) without a boundary between them.The surface mesh with ventricles can be constructed from the segmentation of computed tomography (CT), magnetic resonance imaging (MRI), or echocardiography data.Our automated algorithm was tested on 400 triangular surfaces with LV and RV obtained from semi-automated CT segmentation.The mean time for modeling was 87 seconds.As a result, our algorithm worked successfully in 97% of the cases.In addition, we validated the results on a manually generated dataset with a 17-segment bull's eye model.For basal, mid, and apex segments we obtained a Dice coefficient of 0.86±0.05,0.76±0.09,and 0.56±0.17respectively.
Source: Computing in cardiology
Institution: Federal Almazov North-West Medical Research Centre
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4390446972