Today, centers and
clinics often collaborate to enable imaging studies involving hundreds and even
thousands of human test subjects. In preclinical imaging research however, such
multi-institution studies are much rarer, despite the fact that in this setting
too, the pooling of larger amounts of experimental data would improve the
statistical validity of obtained results. Of course, that statistical validity
would depend also on the standardization of acquisition and analysis protocols,
if multicenter preclinical imaging studies are to become a reality.
The Processing and Analysis of
Multimodal Images team of MIRCen's Neurodegenerative Diseases
Laboratory participated in a study aimed at demonstrating the feasibility and
interest of collaborative multicenter preclinical imaging research. The raw
magnetic resonance imaging (MRI) data of 40 healthy rat brains produced by two
centers were pooled and analyzed in software developed at four centers. The
study's authors focused on mapping T1 and T2 relaxation times at a magnetic
field strength of 7 Tesla. Relaxometry maps permit the study of brain anatomy,
particularly the contrast between its white and gray matter. By comparing and
analyzing their results, the team was able to quantify observer variability
between the two contributive centers. They also performed an evaluation of the influence
of imaging processing steps on the resulting maps, using three different fitting
algorithms from as many different centers. Finally, the relaxation times were
studied in different brain regions of interest using image analysis solutions
developed by two of the centers. The differences measured between T1 and T2
data were inconsequential.
The study thus showed that multicenter
research is feasible in the preclinical setting. The obtained relaxometry maps
of healthy rat brains may prove useful as references for future works on signal
intensity modifications in experimental pathological models.
Top left: Individual sagittal and coronal T1-weighted raw images (inversion
time: 247 ms) Top right: Individual sagittal and coronal T2-weighted raw images
(echo time: 50 ms). Middle left: Sagittal and coronal views of the mean T1
relaxometry model (n=40 rats) Middle right: Sagittal and coronal views of the
mean T2 relaxometry model (n=40 rats) Bottom left: Sagittal and coronal views of
the T1-weighted model (40 rats; inversion time: 247 ms). Bottom right:
Parcellation of the T1-weighted model (each brain region colored differently)
(taken from the article).
© Frontiers in Neuroinformatics
1 To learn more about the FLI initiative
2 To learn more about SAIN
3 INSERM, U1216, Grenoble Institute of
Neurosciences, Grenoble Alpes University / CNRS, CRMBM, Aix-Marseille
University / UMR 9199, CNRS, CEA-MIRCen, Paris-Saclay University / CNRS,
ICube-IMAGeS, University of Strasbourg