1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Title:	The identification of new CT scan markers to diagnose cerebral amyloid angiopathy in primary intracerebral hemorrhage
Authors:	Nor Shahirah Shaik Amir (P83586)
Supervisor:	Kalaivani Chellappan, Dr.
Keywords:	Tomography Hypertension Image processing Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia
Issue Date:	29-Mar-2019
Description:	Primary intracerebral hemorrhage (p-ICH) most commonly occurs due to chronic hypertension or cerebral amyloid angiopathy (CAA). The gold standard of diagnosing the p-ICH is using Magnetic Resonance Imaging (MRI) that can differentiate between more obvious cases but Computed Tomography (CT) remains the preferred imaging modality for its availability, affordability and accessibility in most developing countries. Differentiating them on CT is the challenge addressed in this study. The problem remains in distinguishing CAA from hypertensive ICH based on CT brain image. We propose to improve the yield of CT imaging to better differentiate between the two. In this study, 2-D CT brain images (n=40; 30 males, 10 females) of p-ICH were acquired from Universiti Kebangsaan Malaysia Medical Centre (UKMMC), Imaging Centre Patient Database. The processing of the images was all done in Matlab® environment. The CT brain images were obtained in digital imaging and communications in medicine (DICOM) format. Anonymization was implemented to remove the metadata from the images for further processing. CT image preprocessing is the first module that consist of image resampling was applied to resize the images from 512×512 into 256×256 pixel resolutions using the Lanczos algorithm to reduce the aliasing effect on the edges of the image with MSE value of 204.86. The images were then normalized to standardize the pixel intensity from range of [0 65535] into [0 255] for a clearer image. In the next section, the noise present in all 40 CT images were reduced and compared with wiener filter, non-linear means and wavelet method. Wiener filter outperforms others with the average MSE and PSNR values of 206.26 and 25.02. The denoised images were enhanced in the next step by applying modified unsharp masking algorithm and compared with conventional unsharp masking algorithm. The proposed method achieved a better result with average values of MSE, PSNR and SSIM of 75.16, 29.66 and 0.97. The following module segments out the skull and extracts the brain for further CAA analysis utilizing the thresholding technique, image labeling and morphological operation. Thresholding with the average mean value of 0.90 was used to identify the skull and visible hemorrhage present in all the images. Image labeling was done to select the skull and in the final segment, morphological operation was applied to remove the skull from the CT brain image. The identification of CAA hemorrhage from h-ICH was improved through the k-means algorithm. It identifies 7 clusters, namely; 1 cluster of white matter and 2 clusters of gray matter, cerebrospinal fluid and hematoma. CT clustered images were validated with a neurologist and the inter-reliability test was measured using Cohen Kappa. Kappa value obtained was 0.88 which indicates a strong level of agreement between the developed algorithm and the neurologist. Verification was done between the algorithm against the corresponding MR images and it achieved accuracy of 86%, sensitivity of 100% and specificity of 67%. Location based identification is the most significant indicator in differentiating CAA (lobar region) and h-ICH (non-lobar region) in this study.,Master of Science
Pages:	104
Call Number:	RC78.7.T6N637 2019 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.