Abstract—This paper provides the explanation of the concepts of similarity measuring technique to be implemented in image registration process. Template matching has been used for many applications in image processing. This paper discussed about the implementation of template matching for automatically detection of the mitral valve image. An experiment is carried out which covers the patient scanning who suffers from mitral valve disease. Here, the normalized cross correlation is employed to conduct the template matching. The performance of the method is validated by comparing the value of correlation coefficient produces by different images.
Keywords—Template matching, Mitral valve, Image processing, Normalized cross correlation.
This work is supported by Graduate Research Grant (GPPS: no of U955).The authors are with the Department of Electrical and Electronic Engineering and Department of Power Electrical Engineering, Faculty of Electrical and Electronic Engineering, University Tun Hussein Onn Malaysia (UTHM), 86400, Parit Raja, Batu Pahat, Johor, Malaysia
The two chambers (atrium and ventricle) of the left side of the heart are separated by mitral valve. Mitral valve prolapse is the condition when the leaflets of the mitral valve bulge (prolapse) into the left atrium as the heart contract 1. Mitral valve regurgitation or incompetence is a heart disorder when the mitral valve does not properly closed when the heart pumps the bloods out, thus, the blood leaking into the atrium from the ventricle causes by mitral valve prolapse 2. Therefore, during the contraction, it allows blood to flow in two directions. Supposedly, the blood will flow from the ventricle through the aortic valve, however, in this case some of the blood will flow back into the atrium. Thus, the blood volume and pressure in the area will increase due to the leakage.
The evaluation of patients with mitral valve disease is one of the most challenging and promising clinical application of echocardiography. The purposes of this research is to deal with echocardiographic assessment of the mitral valve. Ultrasound machines are non-invasive, low cost and save time, but it depends mostly on the operators skills and experiences to capture clear images. Image processing is vital which involve altering the features of an image in order to either better its descriptive information for human interpretation or render it more suitable for autonomous machine recognition.
Image registration is a process to find corresponding points in two images or more either by automatic or manual procedure and align them into a single image. Template matching is one of the methods used to find the area of a sub image (template) that match to the reference image. This paper implements the idea of the template matching on mitral valve image by using normalized cross-correlation method.
Theory and related work
The mitral valve, otherwise called the bicuspid valve or left atrioventricular valve, is a valve with two folds in the heart, which lies between the two chambers (left atrium and the left ventricle). The mitral valve and the tricuspid valve are known collectively as the atrioventricular valves because they lie between the atria and the ventricles of the heart. Fig. 1 (a) shows the view of opened heart where white arrow indicates of normal blood flow and mitral valve labelled at the center right.
In typical conditions, blood courses through an open mitral valve amid diastole with compression of the left atrium, and the mitral valve closes amid systole with constriction of the left ventricle. The valve opens and closes as a result of pressure contrasts, opening when there is more prominent pressure in the left atrium than ventricle, and shutting when there is more noteworthy pressure in the ventricle than atrium. In anomalous conditions, blood may stream in reverse through the valve (mitral regurgitation) or the mitral valve might be limited (mitral stenosis). Rheumatic heart disease often affects the mitral valve; the valve may also prolapse with age, and be affected by infective endocarditis.
The mitral annulus as shown in Fig. 1 (b) is a fibrous ring that is attached to the mitral valve leaflets. Unlike prosthetic valve (artificial), it is not continuous. The mitral annulus is saddle shaped and changes in shape throughout the cardiac cycle 3. The annulus contracts and reduces its surface area during systole to help provide complete closure of the leaflets. Otherwise, if the leaflets does not close tightly it can result to functional mitral regurgitation 4. The normal diameter of the mitral annulus is 2.7 to 3.5 centimeters, and the circumference is 8 to 9 centimeters. Microscopically, there is no evidence of an annular structure anteriorly, where the mitral valve leaflet is contiguous with the posterior aortic root 5.
Fig. 1. (a) View of the open heart (b) Mitral annulus
The idea of the template matching is to compare and match the rectangular blocks of pixels from a set of two M ×N images of the template and the selected image. The template must be smaller that the reference image for this to work. In this process, the template is being moved pixel by pixel on the selected image to obtain the most matched pixels. Fig. 2 shows the template matching A