Development Of Adaptive Segmentation Techniques For Analysis Of Medical Images

Lizawati, Salahuddin and Norhidayah, Mohamad Yatim and Siti Amaniah, Mohd Chachuli (2012) Development Of Adaptive Segmentation Techniques For Analysis Of Medical Images. Project Report. UTeM, Melaka, Malaysia. (Submitted)

PDF (Full Text) Development_Of_Adaptive_Segmentation_Techniques_For_Analtsis_Of_Medical_Images.pdf - Submitted Version Restricted to Registered users only Download (3MB)

Abstract

This project presents brain lesion segmentation of diffusion-weighted magnetic resonance images (DWI) based on thresholding technique and gray level co-occurrence matrix (GLCM). The lesions are hyperintense lesion from tumor, acute infarction, haemorrhage and abscess, and hypointense lesion from chronic infarction and haemorrhage. Pre-processing is applied to the DWI for intensity normalization, background removal and intensity enhancement. Then, the lesions are segmented by using two different methods which are thresholding technique and GLCM. For the thresholding technique, image histogram is calculated at each region to find the maximum number of pixels for each intensity level. The optimal threshold is determined by comparing normal and lesion regions. Conversely, GLCM is computed to segment the lesions. Different peaks from the GLCM cross-section indicate the present of normal brain region, cerebral spinal fluid (CSF), hyperintense or hypointense lesions. Minimum and maximum threshold values are computed from the GLCM cross-section. Region and boundary information from the GLCM are introduced as the statistical features for segmentation of hyperintense and hypointense lesions. The proposed technique has been validated by using area overlap (AO), false positive rate (FPR), false negative rate (FNR), misclassified area (MA), mean absolute percentage error (MAPE) and pixels absolute error ratio (rerr). The results are demonstrated in three indexes MA, MAPE and rerr, where 0.3167, 0.1440 and 0.0205 for GLCM, while 0.3211, 0.1524 and 0.0377 for thresholding technique. Overall, GLCM provides better segmentation performance compared to thresholding technique.

Actions (login required)

View Item