Françoise Durand-Dubief
The primary objective of our work was to study the loss of the overall brain volume, ie, brain atrophy, as a potential marker of neurodegeneration in patients with multiple sclerosis (MS). Our hypothesis was that overall brain atrophy, a possible index peculiar to each patient, would be predictive of long-term disability. This hypothesis was the consequence of Prof. Confavreux’s clinical and epidemiologic studies on disability accumulation in MS.1,2
Since 2004 we have set up a cohort of 90 patients with MS with a biannual MRI follow-up. The postprocessing of individual MRI data showed an important variability in the procedures of measurement of inter- and intraobserver variability, or inter-MRI scan-rescan. We decided, therefore, to redirect our study towards the assessment of the variability of the measurement of overall brain atrophy using several postprocessing algorithms. Our work aimed at assessing the robustness of 7 different postprocessing algorithms applied to images acquired from different MR imaging systems with data from 9 patients with MS who were followed longitudinally over 1 year (3 time points) on 2 MR imaging systems. Brain volume change measures were assessed using 7 segmentation algorithms: a segmentation-classification algorithm, FreeSurfer, BBSI, KN BSI, SIENA, SIENAX, and JI.
Our work provides further elements for the analysis of factors interfering with the measurement of brain atrophy and for the comparison of various postprocessing algorithms. Identifying these confounding factors remains difficult as there is no way of having a reliable measure of the studied brain volumes and, consequently, of brain atrophy. Several other research teams worldwide have also studied this issue. To date, measuring overall brain atrophy remains a technical challenge. It should be restricted to specialized research teams, and the results should be used with caution.