In this work we aim at investigating the 3D Simple Harmonic Oscillator based Reconstruction and Estimation3 (3D-SHORE) derived numerical indices for quantitative tractography. In particular, we target the cortical motor network (SC-MN) of a cohort of ten healthy subjects. Using diffusion spectrum imaging (DSI) we reconstructed the network connections and compared the resulting information about white matter (WM) density and structure to that provided by Generalized Fractional Anisotropy (GFA) and Magnetization Transfer Ratio (MTR). The SC-MN gathers the connections between the cortical motor area, the basal ganglia and the thalamus, and it essentially consists of three major subcortical networks (Figure 1).

Tractometry of the subcortical motor network using SHORE-based indices

‡ S. Obertino, ‡ M. Zucchelli, † A. Daducci, ┴ C. Granziera*, ‡ Gloria Menegaz* ‡  Dept.  of  Computer  Science,  University  of  Verona.  

†  École  Polytechnique  Fédérale  de  Lausanne.  ┴  Centre  hospitalier  universitaire  Vaudois.  *  equal  contribuEon  

gloria.menegaz@univr.it, sivy.obe@gmail.com  

 Introduc*on  

 Methods  

Ten healthy subjects (age 56.1±17.8 years old, mean±SD) went through a DSI scan twice one month apart (± 1 week, tp1c and tp2c, see [1] for more details). The Ensemble Average Propagator (EAP) was reconstructed using the SHORE model [6] and the orientational (ODF) and microstructural indices were derived including Return to zero (RTOP), Return to axis (RTAP) and Return to plan (RTPP) probability and propagator anisotropy (PA) as in [6]. From RTAP, an estimation of the mean ensemble value of the axons’ radius (R) can be inferred [5]. Tractometry was performed as in [2] and SHORE indices were extracted for each fiber bundle. The repeatability of the measurements across time-points was defined in terms of the stability of the respective probability density functions across time points. This was assessed by measuring the percent absolute changes across time of parametric histogram features (mean, variance, skewness, kurtosis and peak hight) and by calculating the distances between histograms (Mean Square Error, Kullbach-Leibler, Hausdorff). The reported measures represent the ensemble values across region pairs and subjects.

 Results  and  Discussion  

References. [1] S. Obertino, et al, ISMRM Workshop 2013. [2] C. Granziera, et al, Neurology, 2012. [3] P. Hagmann, et al, PLoS ONE, 2007. [4] J. Wedeen, et al, NeuroImage, 2008. [5] A. Daducci, et al, PLoS ONE, 2012. [6] E. Ozarslan, et al, NeuroImage, 2013. [7] F. Aboitz, et al, Brain Res.,1992. [8] D. Alexander, et al, NeuroImage, 2010.

Our results confirm that all the indices are reproducible over time in the SC-MN of normal subjects. Last, our data show that 3D-SHORE indices are highly correlated to anisotropy measures like GFA and much less with measures of myelin presence (i.e. MTR) with the exception of R. This might suggests that the applied 3D-SHORE can be sensitive enough to detect the effect of myelin on axonal diameter.

Figure  2  

Table  1  

Figure  1  Subcor*cal  Networks:  

M1        =    primary  motor  area  PMd  =    premotor  dorsal  area  PMv    =    premotor  ventral  area  SMA    =    supplementary  motor  area  GPi        =    globus  pallidus  interna  

Caudatus  Putamen  

Globus  Pallidus  

Thalamus  

Premotor  loop  

Thalamus  

Putamen  

Caudatus  

GPi  

vPM/dPM  

SMA  loop  

Thalamus  

Putamen  

Caudatus  

GPi  

SMA  

Sensory  Motor  loop  

M1   Thalamus  

Putamen   GPi  

 Conclusions  

Table  2  

% Mean Variance Skewness Kurtosis Peak height MSE (e-5) KLD (e-2) Hausdorff (e-2) GFA 4±1 7±2 17±5 21±6 14±3 3,0±2,9 7,2±6,5 5,0±2,1

MTR 5±1 46±14 13±4 22±5 20±4 46,3±44,1 40,9±34,1 19,2±9,8

RTAP 4±2 9±2 8±10 13±24 19±5 7,2±13,8 6,5±7,3 6,8±5,2

Radius 2±1 11±5 17±9 31±25 15±3 4,5±4,3 8,1±7,7 6,2±2,6

PA 4±1 8±2 12±5 20±9 14±5 3,7±3,5 6,1±5,5 5,6±2,4

GFA-RTAP GFA-R GFA-PA GFA-MTR MTR-RTAP MTR-R MTR-PA Pearson Spearman Pearson Spearman Pearson Spearman Pearson Spearman Pearson Spearman Pearson Spearman Pearson Spearman

Mean 86,44 48,18 86,25 47,27 87,67 81,82 10 22,73 -11,15 -12,73 -27,08 -21,82 21,82 17,27

MSE 55,18 32,73 76,19 42,73 97,79 85,46 39,25 27,27 -26,93 -41,82 68,64 59,09 36,98 23,64

KLD 81,89 70 93,85 88,18 97,51 95,46 66,17 40,91 23,75 10 49 22,73 61,98 32,73

Hausdorff 49,07 40,91 79,46 52,73 96,45 84,55 46,44 34,55 -36,71 -40 71,58 78,18 39,45 35,46

Comparative analysis revealed that 1)  the indicies are not normally distributed along the bundles for any

pair of regions (Jarque-Bera, p<0.001); 2)  histograms are stable across time points for all indices (GFA,

MTR,RTAP, R, PA) (Table 1); 3)  RTAP, R and PA absolute percent changes on mean is highly

correlated with those of GFA as well as histogram distances; 4)  PA has the highest correlation with GFA, while R has the highest

correlation with MTR and is highly correlated with GFA; 5)  the distributions of the estimated axons radius is consistent across

the fiber bundles and the values reported in literature [7,8] (Figure 2: average distribution across subjects).