Robust and efficient methods for segmentation of intima media thickness of the common carotid artery

Abstract

Cardiovascular diseases are the third leading cause of death worldwide. The primitive indication of the possible onset of a cardiovascular disease is atherosclerosis, which is the accumulation of plaque on the arterial wall. To assess carotid atherosclerosis, non invasive ultrasound imaging modality is preferred over other invasive methods due to their safer profile and ability to explore atherosclerosis in its early stages. The intima media thickness (IMT) of the common carotid artery (CCA) is an early marker of the development of cardiovascular disease. The computation of IMT and the delineation of carotid plaque are significant predictors for the clinical diagnosis of the risk of stroke. However, manual measurement of the IMT is tedious, error-prone and subjected to observer variability. Hence, there is a growing interest in the development of automated software system for the measurement of IMT from the carotid ultrasound images. The development of such automated systems is the primary objective of this research. The presence of speckle noise in carotid ultrasound image reduces the quality of image and automatic human interpretation. Carotid ultrasound images have multiplicative speckle noise and it is difficult to remove as compared to the additive noises. The despeckling filters have a greater restriction on preservation of edges and characteristics. For a robust diagnosis, carotid ultrasound images must be free from speckle noise. To address this problem, we propose the use of a Bayesian least square estimation method for the reduction of speckle noise in logarithmic space. In addition, one of the widely accepted method named optimized Bayesian non local mean filter is adopted in our work to reduce the speckle noise in ultrasound images. The traditional denoising techniques require a significant amount of execution time because of the iterative steps involved. To overcome this problem, we propose the use of Wiener filtering in the wavelet domain. Wiener filter smoothens the image while retaining the edges, and performs region of interest (ROI) extraction significantly faster than other similar techniques. Further, the state-of-the-art enhancement techniques are adopted in order to increase the contrast after denoising. Finally, the comparative study of edge detection algorithms is done based on the framework of despeckling carotid ultrasound images. iiiIn the literature, several edge-based algorithms is proposed for estimating the IMT. However, accurate segmentation still remains a challenge. Extracting the ROI prior to the segmentation from carotid ultrasound images has been very much challenging as it is the basis for further image analysis, interpretation, and classification. In order to extract ROI of ultrasound images, several types of morphological functions are applied. The identified region is analysed to detect the carotid wall boundaries. The ROI extraction must be performed properly otherwise it leads to a lot of misinterpretation, and false measurement. To address this problem, we present novel approaches for automatic ROI extraction followed by new segmentation algorithms based on threshold-based wind driven optimization, support vector machine and structured random forest classifier for measurement of the IMT of the CCA. The results obtained are compared with the state-of-the-art algorithms, and the results show that the proposed methods outperform the existing techniques in terms of IMT segmentation accuracy and computational speed.