2020 Volume 38 Issue 1 Pages 21-26
Recent progress in image analysis methods including a deep learning open a new phase in the analysis using echo-cardiographic images. Indeed, a disease-existence diagnosis, disease-difference diagnosis, and functional diagnosis via the computational analysis with echocardiographic images are nowadays the active research field. In this paper, we introduce the research of cardiac conditions using two-dimensional echo imaging, especially, a quantitative evaluation of left ventricular ejection fraction (LVEF). Although the previous LVEF evaluation is needed through an observer dependent process that requires manual tracing, we show here that the direct evaluation from echocardiographic images using deep neural network with various view modes gives precise prediction in LVEF. In addition, the pre-processing used in our research, including automated classification of imaging plane, is introduced.
