Automated detection of aortic annulus sizing based on decision level fusion

Abstract

Aortic valve disease occurs due to calcification on the area of leaflets and it is progressive over time. Surgical Aortic Valve Replacement (SAVR) can be performed to treat the patient. However, due to invasive procedure of SAVR, a new method known as Transcatheter Aortic Valve Implantation (TAVI) has been introduced, where a synthetic catheter is placed within the patient’s heart valve. Traditionally, aortic annulus sizing procedure requires manual measurement of scanned images acquired from different imaging modalities which are Computed Tomographic (CT) and echocardiogram where both of the modalities produce inconsistency in measuring the aortic annulus yet able to produce different parameters which lead to accurate measurement. In this research, the image processing techniques of CT scan and echocardiogram images are done separately in order to obtain the aortic annulus size. Intensity adjustment and median filter are applied to CT scan image pre-processing, Watershed Transformation associated with the morphological operation has been used to perform the aortic annulus segmentation while image resizing and wavelet denoising method have been performed in echocardiogram image pre-processing followed by the implementation of Otsu N-clustering and morphological operation method for object segmentation. Then, Euclidean distance formula is applied to measure the distance between two points that indicates the diameter of the aortic annulus. Finally, a decision fusion technique based on the mathematical statistic approach has been applied to fuse the measured annulus size obtained from both modalities. Results affirmed the approach’s ability to achieve accurate annulus measurements when the final results are compared with the ground truth. In addition, the application of non-probabilistic estimation on the decision level fusion approach which does not required the dataset training produces fast computational time and helps in determining the optimal size of new aortic valve to be implemented in human heart.