2019 Research Forum

Department of Radiology

Applicant: Ronald Crandall MS IV Principal Investigator & Faculty Sponsor: Sudha Challa MD Automated and Standardized Quantification of Mild Cerebral Small Vessel Disease in Computed Tomography Ronald Crandall MS IV, Sudha Challa MD

INTRODUCTION The standardized quantification and interpretation of mild white matter change (WMC) in cerebral small vessel disease have been debated. Magnetic resonance imaging (MRI) has been used as the gold standard for WMCs. However, computed tomography (CT) is widely used and a cheaper imaging modality and therefore an algorithm that identifies WMCs would be valuable. WMCs on CT are difficult to detect given the narrow range of densities in brain parenchyma. Automated means may provide a more standardized detection for WMCs on CT. PURPOSE The aim of this study is to demonstrate that automated methods can result in better identification and early detection of white matter changes on CT. METHOD We performed a retrospective analysis of Head CT images utilizing two software packages, the Analysis Group at the Oxford Centre for Functional MRI of the Brain Software Library (FSL) and open source Neuroconductor R (R = statistical software). FSL is used by nearly 1000 hospitals and university labs and has received over 2500 citations and was validated for CT brain extraction by Muschelli et al. and CT tissue segmentation by Cauley et al. Our modified algorithm performs automated brain extraction, tissue-type differentiation, and visually highlights hypodense regions to identify WMC. These hypodense regions were subsequently analyzed. Initial algorithm training was completed using 6 CT studies and corresponding MRIs to identify mild to severe WMC. The results were correlated to previously reported findings and normal age-matched controls to validate the proposed algorithm. RESULTS A total of 32 patients have been studied so far. When compared with true WMCs validated by MRI findings (6 patients), the initial data suggest a significant change in the WM volume and density. While the periventricular WM density in the 3rd decade group approached significance at (p=0.19) it was not found to be significant in the 4th decade group. However, the total WM density in the 4th decade group approached significance at (p=0.16) using a one-tailed test. Age Group Total WM Volume* Total WM Density Near CSF WMC volume Distant from CSF WMC volume 20-29 normal (10) 633.52 32.83 1.95 1.26 20-29 with WMCs (5) 558.41 32.31 7.04 1.55 30-39 normal (4) 624.84 33.02 3.13 2.05 30-39 with WMCs (7) 645.02 31.88 4.50 2.84 * volume is in cm3, density is in Hounsfield Units DISCUSSION Even though CT images are less sensitive than MRI, CTs are utilized more frequently, are performed more quickly and are less costly. Mild WMCs on CT are difficult to interpret and in practice are not graded with the Age-Related White Matter Change scale by Wahlund, et al. In our group of patients in the 4th decade with WMCs while total WM volume increased unexpectedly, our algorithm correctly identified decreased total WM density and increased WMC volumes. This algorithm may allow CT to be utilized to correctly identify WMCs and volumes over time. CONCLUSION Clinical use of automated detection of WMCs will help standardize initial disease stage per decade and progression. For further research, it will be interesting to apply the above techniques to a larger population for training and study with non-trauma and non-ischemia related diseases. This may help foster preventative measures.

126