Quantification of Thalamic Volume in Multiple Sclerosis: From the Multicenter INNI Dataset Towards the Clinical Application
Loredana Storelli1, Elisabetta Pagani1, Patrizia Pantano2, Gioacchino Tedeschi3, Nicola De Stefano4, Paolo Preziosa5, Maria Rocca5, Massimo Filippi6
1Neuroimaging Research Unit, Division of Neuroscience, IRCCS San Raffaele Scientific Institute, 2Department of Human Neurosciences, Sapienza University of Rome; and IRCCS NEUROMED, 3Department of Advanced Medical and Surgical Sciences, and 3T MRI-Center, University of Campania "L. Vanvitelli", 4Department of Medicine, Surgery and Neuroscience, University of Siena, 5Neuroimaging Research Unit, Division of Neuroscience, and Neurology Unit, 6Neuroimaging Research Unit, Division of Neuroscience, Neurology Unit, Neurorehabilitation Unit, and Neurophysiology Service, IRCCS San Raffaele Scientific Institute; Vita-Salute San Raffaele Unviersity
Objective:
To obtain a reliable segmentation of the thalamus in multiple sclerosis (MS) by comparing existing automatic methods.
Background:
Thalamic atrophy has been found since the earliest phases of MS. However, this measure is not included in clinical practice, due to the time-consuming manual segmentation and technical challenges.
Design/Methods:
141 relapsing-remitting MS and 69 healthy controls (HC) with baseline and 1-year 3D T1-weighted, T2-weighted and diffusion weighted (DW) MRI were collected from the Italian Neuroimaging Network Initiative. From DWI, fractional anisotropy (FA) maps were derived to be used with T1-weighted MRI, as input for the FSL-MIST multimodal segmentation. FSL-FIRST v5.0.9 and Freesurfer v6.0 were also compared, both at baseline and at follow-up. The agreement among the results of the pipelines and the effect sizes in differentiating between MS and HC were assessed. In patients, correlations with age, disease duration, EDSS and T2-hyperintense lesion volume (LV) were evaluated.
Results:
At baseline, FIRST and MIST showed the highest significant agreement in the results of thalamic volume (R=0.87, p<0.001), with the highest effect size for MIST (Cohen’s d=1.11). At baseline, FIRST showed the highest significant correlations with age (-0.36, p<0.001), EDSS (R=-0.3, p<0.001, adjusted for age), T2-hyperintense LV (R=-0.4, p<0.001) and disease duration (R=-0.2, p=0.02). At follow-up, MIST showed the lowest variability in estimating thalamic volume changes (TVC) for HC (standard deviation=1.07%) in comparison to the other pipelines, and the highest effect size (Cohen’s d=0.21). In MS patients, only MIST TVC showed a significant correlation (adjusted for age) with T2-hyperintense LV change (R=-0.22, p=0.01).
Conclusions:
We found that the inclusion of FA contrast increased robustness of the longitudinal results and a better capability to detect small variations of thalamic volumes, as shown by MIST results. The advantage of a multimodal approach is also shown by the results of correlations with LV changes for MIST.