Validation of Routine High-Resolution 3T MRI Morphometry in Multiple Sclerosis (P6.164)

OBJECTIVE: Assess the accuracy of novel automated multi-channel high-resolution MRI morphometry in multiple sclerosis (MS) for long-term routine assessment from clinically suitable 3T high-resolution images. BACKGROUND: We have developed new software for the automated segmentation and measurement of whole brain lesion burden and atrophy from high-resolution MRI images in MS. The system is geared toward routine assessment, i.e. optimized for consistency in long-term follow-up and across a wide spectrum of disease burden. METHODS: The MRI protocol (3T Siemens Skyra, 20-channel headcoil) comprised three 1mm isotropic scans with T1, T2 and fluid-attenuated inversion-recovery (optimized for depicting white/gray matter, CSF, and WM lesions (WML), respectively. The automated segmentation pipeline included co-registration, distortion correction, local and global intensity normalization, and heuristics to correct for common tissue misclassifications, providing better measurement stability in the presence of long-term scanner drift and protocol and hardware updates, which are inevitable in long-term clinical routine imaging. Validation of accuracy occurred against repeated manual tracing of WML and CSF in 14 subjects selected for a wide range of disease burden. The patients (n=14) were a mix of relapsing (n=9) and progressive (n=5) forms of MS and had a (mean±SD) age 51.1±12.0 years, disease duration 13.1±7.1 years, and Expanded Disability Status Scale score 2.5±1.8. RESULTS: Measured WML volume was a range of 0.5-41 ml. Intra-rater agreement for manual segmentation was 0.998 and 0.970 for WML and CSF volumes (intra-class correlation, ICC), respectively. The ICC agreement between the expert (manual) and fully automated segmentation for WML and CSF volume was 0.965 and 0.870, respectively. CONCLUSIONS: A high level of agreement between uncorrected, fully automated morphometry and manual expert tracing was observed across a wide spectrum of disease burden, suggesting that accurate high-resolution MS morphometry is possible at 3T MRI via uncorrected automated segmentation of a 3-channel 3T MRI protocol. Disclosure: Dr. Tummala has nothing to disclose. Dr. Khalid has nothing to disclose. Dr. Moscufo has nothing to disclose. Dr. Cavallari has nothing to disclose. Dr. Bakshi has received personal compensation for activities with Biogen Idec, Novartis, Sanofi/Genzyme, and Teva Neuroscience as a consultant. Dr. Guttman has stock and/or stock options in Novartis, Roche, GlaxoSmithKline, Alnylam, Cocrystal Pharma, Arrowhead Research, Sangamo BioSciences, Inc., and Protalix Biotherapeutics. Dr. Weiner has received personal compensation for activities with Biogen Idec, Novartis, EMD Serono, and Teva Neuroscience. Dr. Meier has nothing to disclose.