Fisher, Jason2016-02-102016-02-102014https://hdl.handle.net/2144/14368OBJECTIVE: Determine if complex singular value decomposition (cSVD) used as preprocessing in the maximum slope algorithm reduces image noise of resultant physiologic parametric images. Noise will be decreased in the parametric maps of cerebral blood flow (CBF), cerebral blood volume (CBV) as compared to the same algorithm and data set with no cSVD applied. MATERIALS AND METHODS: A set of 10 patients (n=15) underwent a total combined 15 CT perfusion studies upon presenting with stroke symptoms. It was determined these patients suffered from occlusions resulting in a prolonged arrival time of blood to the brain. DICOM data files of these patients scans were selected based on this increased arrival delay. We compared the output of estimation calculations for cerebral blood flow (CBF), and cerebral blood volume (CBV), using preprocessing cSVD against the same scan data with no preprocessing cSVD. Image noise was assessed through the calculation of the standard deviation within specific regions of interest copied to specific areas of grey and white matter as well as CSF space. A decrease in the standard deviation values will indicate improvement in the noise level of the resultant images.. Results for the mean value within the regions of interest are expected to be similar between the groups calculated using cSVD and those calculated under the standard method. This will indicate the presence of minimal bias. RESULTS: Between groups of the standard processing method and the cSVD method standard deviation (SD) reductions were seen in both CBF and CBV values across all three ROIs. In grey matter measures of CBV, SD was reduced an average of 0.0034 mL/100g while measures of CBF saw SD reduced by an average of 0.073 mL/100g/min. In samples of white matter, standard deviations of CBV values were reduced on average by 0.0041mL/100g while CBF SD's were reduced by 0.073 mL/100g/min. CSF ROIs in CBV calculations saw SD reductions averaging 0.0047 mL/100g and reductions of 0.074 mL/100g/min in measures of CBF. Bias within CBV calculations was at most minimal as determined by no significant changes in mean calculated values. Calculations of CBF saw large downward bias in the mean values. CONCLUSIONS: The application of the cSVD method to preprocessing of CT perfusion imaging studies produces an effective method of noise reduction. In calculations of CBV, cSVD noise reduction results in overall improvement. In calculations of CBF, cSVD, while effective in noise reduction, caused mean values to be statistically lower than the standard method. It should be noted that there is currently no evaluation of which values can be considered more accurate physiologically. Simulations of the effect of noise on CBF showed a positive correlation suggesting that the CBF algorithm itself is sensitive to the level of noise.en-USMedical imagingCTImagingPerfusionProcessingRadiologyStrokeThesis/Dissertation2016-01-22