Feasibility Study of Echocardiographic Images Segmentation Based on Sparse Representation

• Published in: Iranian journal of electrical & electronic engineering Vol. 18; no. 2; p. 1993
• Main Authors: S. Fouladifard, H. Behnam, P. Gifani, M. Shojaeifard
• Format: Journal Article
• Language: English
• Published: Iran University of Science and Technology 01.06.2022
• Subjects: dictionary design; echocardiography; intensity variation time curves (ivtc) signal; sparse representation; temporal super-resolution
• ISSN: 1735-2827; 2383-3890

A semi-automatic method for the segmentation of the Left Ventricle in echocardiography images is presented. The manual segmentation of the left ventricle in all image sequences takes a lot of time. The proposed method is based on sparse representation and the design of overcomplete dictionaries based on prior knowledge of the intensity variation time curves (IVTC). We used the sparse recovery algorithm of orthogonal matching pursuit (OMP) to find the sparse coefficients of the IVTC signals. We obtained the histogram of non-zero sparse coefficients for all images. The binary images from successive frames were constructed via thresholding. In addition, we defined one image representing all the frames, dividing all the points of the heart into three groups. One group involved the points located inside the cavities in all frames. The second group included the points that belonged to the tissue in all frames. Points that in some frames are located inside the cavities and in some other frames are located inside the tissue. The results on 2D echocardiographic images acquired from both healthy and patient subjects showed good agreement with manual tracing and took a short time for the contour, including the whole left ventricle. According to the cardiology specialist, the value of ejection fraction is correctly calculated, and the error percentages were 0.83 and 2.33 for two healthy data samples. The proposed method can be applied to 3D echocardiography images to obtain the left ventricular volume. This approach also can be used for other types of medical images.