Related Experiment Video
Updated: Feb 27, 2026
09:41
Comprehensive Autopsy Program for Individuals with Multiple Sclerosis
Published on: July 19, 2019
12.1K
Volumetric Analysis from a Harmonized Multisite Brain MRI Study of a Single Subject with Multiple Sclerosis
R T Shinohara1, J Oh2,3, G Nair4
1From the Departments of Biostatistics and Epidemiology (R.T.S., K.A.L.) rshi@mail.med.upenn.edu.
AJNR. American Journal of Neuroradiology
|June 24, 2017
Summary
Multicenter multiple sclerosis (MS) brain imaging studies show significant site-related biases in lesion volume and brain atrophy measurements, even with harmonized protocols. These systematic differences can severely impact volumetric analyses and research findings.
Area of Science:
- Neuroimaging
- Radiology
- Multiple Sclerosis Research
Background:
- Magnetic Resonance (MR) imaging is crucial for diagnosing and monitoring multiple sclerosis (MS).
- Standardized imaging protocols are essential for multicenter studies to ensure data consistency.
- The North American Imaging in Multiple Sclerosis Cooperative developed a uniform 3T MR imaging protocol.
Purpose of the Study:
- To assess intersite variability in brain scan data acquired using a standardized MR imaging protocol.
- To evaluate potential systematic biases in lesion quantification and volumetric measurements across different imaging sites.
- To determine the impact of site differences on the reliability of MR imaging data in multiple sclerosis research.
Main Methods:
- MR imaging was performed on Siemens scanners (Skyra, Tim Trio, Verio) at seven different sites.
- Expert manual segmentation was used for T1-hypointense and T2 (FLAIR) hyperintense lesions.
- Automated pipelines were applied for lesion detection and brain volumetric measurements (whole-brain, cortical, deep gray matter).
Main Results:
- Significant site-related biases were observed in both manual T1 and T2 lesion volume measurements (P < .01).
- Site differences explained over 90% of the variation in lesion volumes and over 80% in automated volumetric measurements.
- Imaging output measures showed clustering based on scanner models, with Skyra scanners yielding similar results.
Conclusions:
- Systematic differences between imaging sites can introduce severe biases in volumetric analyses.
- Protocol harmonization and consistent scanner field strength are insufficient to eliminate site-related variability.
- Careful consideration of intersite variability is critical for the interpretation of multicenter neuroimaging studies in multiple sclerosis.
