Improvement of basal ganglia detectability in NCAT Phantom Brain SPECT by wavelet Transformation in Image Processing domain Improvement of basal ganglia detectability in NCAT Phantom Brain SPECT by wavelet Transformation in Image Processing domain

Abstract

The possibility of managing Parkinson's disease is very high if it is diagnosed at the early stage. This study was conducted to improve the quality of radioisotope images in order to increase the ability to detect the basal nuclei of the human brain phantom SPECT in the field of image processing in the MATLAB software environment. Material And Methods: Human brain simulation was done using XCAT digital brain phantom with different activities of Caudate and Putamen nuclei. Then, SPECT scan simulation of the brain was performed by SIMIND Monte Carlo simulated SPECT system according to the SPECT system of Tabriz Imam Reza Hospital. The obtained projections were reconstructed by Iterative reconstruction, and a brain image with normal absorption of the basal nuclei was selected as the input of the image processing using the Wavelet transformation. Denoising was done on the input image through 9 wavelet methods at different levels, and then segmentation was done through 6 methods. The output images were interpereted and selected by a nuclear medicine physicain from different levels of wavelet methods and Adaptive Threshold segmentation, and were used to calculate the evaluation criteria of Sensitivity, Specificity and Dice coefficient. The calculations were done based on a Ground Truth image that was specified by the physician. Results: Adaptive Threshold segmentation at level 7 of the Biorthogonal method(Sensitivity=94%, Specificity=79%, Dice coefficient=61%), levels 7 of the Coiflet method (Sensitivity=98%, Specificity=78%, Dice coefficient=51%), level 6 of the Daubechies method (Sensitivity=98%, Specificity=78%, Dice coefficient=50%), level 5 of the Haar method (Sensitivity=96%, Specificity=80%, Dice coefficient=55%), level 6 of the Morlet method (Sensitivity=96%, Specificity=81%, Dice coefficient=62%) and level 6 of the Symlet method (Sensitivity=98%, Specificity=78%, Dice coefficient=53%) were identified as the best in suitable detection of the basal ganglia nuclei on the reconstructed images from the brain SPECT scan.