Comparison of three U-Net family architectures for left ventricular myocardial wall automatic segmentation | IEEE Conference Publication | IEEE Xplore

Comparison of three U-Net family architectures for left ventricular myocardial wall automatic segmentation


Abstract:

Left ventricular (LV) segmentation is an important process which can provide quantitative clinical measurements such as volume, wall thickness and ejection fraction. The ...Show More

Abstract:

Left ventricular (LV) segmentation is an important process which can provide quantitative clinical measurements such as volume, wall thickness and ejection fraction. The development of an automatic LV segmentation procedure is a challenging and complicated task mainly due to the variation of the heart shape from patient to patient, especially for those with pathological and physiological changes. In this study, we focus on the implementation, evaluation and comparison of three different Deep Learning architectures of the U-Net family: the custom 2-D U-Net, the ResU-Net++ and the DenseU-Net, in order to segment the LV myocardial wall. Our approach was applied to cardiac CT datasets specifically derived from patients with hypertrophic cardiomyopathy. The results of the models demonstrated high performance in the segmentation process with minor losses. The model revealed a dice score for U-Net, Res-U-net++ and Dense U-Net, 0.81, 0.82 and 0.84, respectively.
Date of Conference: 01-05 November 2021
Date Added to IEEE Xplore: 09 December 2021
ISBN Information:

ISSN Information:

PubMed ID: 34891859
Conference Location: Mexico

I. Introduction

The automatic identification of hypertrophic cardiomyopathy (HCM) using cardiac medical images is an emerging and challenging field. The development of medical imaging technologies provides the capability of early diagnosis and detection of the disease. Computed tomography (CT) among other imaging techniques is preferred for the visualization of the heart left ventricle (LV) and the evaluation of cardiomyopathies [1]. Segmentation and delineation of the left ventricle is a crucial step for the quantification of the morphological and pathological changes, providing important clinical variables, such as ejection fraction, end systolic and diastolic volume, wall thickness, etc. However, for most of the imaging modalities used, the manual segmentation of the heart is labor-intensive and time-consuming for a single subject [2]. Thus, automating the segmentation is highly desirable as it can provide significant contribution both in the clinical and the bioengineering domain.

References

References is not available for this document.