S&M2891 Research Paper of Special Issue
Published: April 4, 2022
Ejection Fraction Measurement and Regional Wall Motion Abnormality Assessment Using Deep-learning Neural Networks in Left Ventriculography [PDF]
Shan-Bin Chan, Yuan-Chun Lai, Wei-Ting Chang, Kuo-Ting Tang, Ming-Shih Huang, Zhih-Cheng Chen, and Yung-Yao Chen
(Received July 4, 2021; Accepted September 30, 2021)
Keywords: ejection fraction, regional wall motion abnormalities, deep learning, neural networks, left ventriculography, semantic segmentation, image classification
In this research, an X-ray flat panel detector is adopted as an image collection sensor for evaluating left ventricular systolic functions. Typically, left ventriculography (LVG) is conducted in the end-diastolic and end-systolic areas by clinicians, which is time-consuming, and the calculated ejection fraction (EF) varies among clinicians. We propose two novel methods for EF measurement and regional wall motion abnormality (RWMA) assessment through LVG. Our methods can automatically segment the end-diastolic and end-systolic areas for clinicians and perform EF measurement and RWMA assessment in real time. Semantic segmentation neural networks were implemented for EF measurement, and image convolution neural networks were implemented in RWMA recognition. LVG images were collected by clinicians, but the data set labeling procedure was not performed by clinicians. This method may reduce the need for medical doctors in the data set labeling procedure. Using the proposed methods, EF measurement and RWMA assessment were performed with high accuracy.Corresponding author: Yung-Yao Chen
This work is licensed under a Creative Commons Attribution 4.0 International License.
Cite this article
Shan-Bin Chan, Yuan-Chun Lai, Wei-Ting Chang, Kuo-Ting Tang, Ming-Shih Huang, Zhih-Cheng Chen, and Yung-Yao Chen, Ejection Fraction Measurement and Regional Wall Motion Abnormality Assessment Using Deep-learning Neural Networks in Left Ventriculography, Sens. Mater., Vol. 34, No. 4, 2022, p. 1287-1296.