Voxel-based morphometry strategies comparison in the study of the mouse brain

DELFINA BRAGGIO; PAULA GONZALEZ; JIMENA BARBEITO; IGNACIO LARRABIDE

Tandil

Simposio; 12th International Symposium on Medical Information Processing and Analysis; 2016

SIPAIM society

Voxel-based morphometry strategies comparison in the study of the mouse brainIntroductionVoxel-based morphometry (VBM) is a technique for neuroimage analysis. The aim of this technique is to detect morphologic differences in the brain anatomy at a global level, based on voxel-wise statistics.Since its description in 1999 [1], many studies used this technique, becoming one of the most popular processes in neuroimage analysis to explore differences between groups. Despite VBM is very common in the study of the human brain, its use for other species is not very extended yet.There are some previous researches studying morphologic differences in mice using this technique. But defining a workflow and identifying the suitable parameters to apply this technique is a challenge, mainly because any variation on them has a great impact on the results of the study.In this work, we have contrasted different VBM strategies comparing mice MRIs that have been subjected to different treatments.Materials and methodsImages used in this study were taken from 23 adult mice from the C57 strain. These animals were divided in three experimental groups: control, low calorie-proteine diet (LCP) and low protein diet (LP). LCP and LP correspond to different exposures of intrauterine growth restriction.We compared four different workflows to analyze their impact on the results of the group comparison.ResultsAs a result of all the workflows, we obtain two thresholded statistical map for each group comparison made: one representing those areas where the control group has a bigger amount of gray matter than the experimental group (C>LP and C>LCP) and one where the amount is smaller (LP>C and LCP>C). Making a total of 4 thresholded map for each pair workflow-smoothing kernel.As a way of comparing the workflows, some measures were taken to contrast the classic VBM (WF1) with the rest of them. The Dice coefficient was calculated between the thresholded maps resulted from the different workflows. This coefficient was calculated for the different levels of smoothing applied. It was considered a significance level of 0.01 for uncorrected p-values.ConclusionsA quantitative comparison between different workflows for the application of voxel-based morphometry was made. The different workflows applied on the data have a great impact on the results of the statistical analysis. The Dice coefficient values remain under significant values for all the comparisons made, varying slightly between different values of smoothing. The choice of a specific workflow must be planed before its application on the data, specially when the size of each experimental group is relatively small.