Improved characterization of brain microstructure is important for image-based methods for diagnosing stroke. We explored the extent to which microstructural maps including Fractional Anisotropy (FA), Generalized Fractional Anisotropy (GFA), and Neurite Orientation Dispersion Density Index (NODDI) detect ipsilateral and contralateral differences in stroke patients as a measure of stroke severity. The difference between hemispheres was correlated with Fugl-Meyer Assessment motor function scores and the results of 16 patient scans reported. Results suggest that the Orientation Dispersion Index (ODI) contains information that could be clinically useful in understanding stroke recovery.
Methods
Acquisition and Clinical Assessment of Stroke Severity
16 data sets of motor ischemic stroke subjects were acquired by DSI with a b-max of 4000 and 203 directions1 on a Siemens 3T Verio scanner using a 32 channel head coil. A simultaneous multi-slice blipped CAIPI sequence2 was used with a slice acceleration factor of three. The 16 scans included 9 baseline scans, 5 scans at 1 month time point, and 2 scans at 3 month time point. Following each scan, a Fugl-Meyer assessment (FM) of motor function was performed with higher scores representing better motor control. The assessment includes scores for Upper Extremity (UE) and Lower Extremity (LE) and a composite which is the sum of the two 3.The highest correlation was found for Slow NODDI ODI sKLDs with 1/FM and 1/UE (Table 1). Consistent with the correlation results, paired t-tests for the mean value in the ipsilateral and contralateral stroke regions showed that only for the NODDI ODI maps were the mean values significantly different between hemispheres (original p=0.00089, AMICO p=0.0016). Correlation with FM for ipsilesional mean values were below 0.3, but correlation of contralesional mean values in FA and RDI was above 0.6 (Table 1). Volume-scaling sKLD values raised the correlation of GFA and CSF microstructure maps 30-60% (Fig 3).
The increase in correlation in the GFA and CSF maps after scaling by volume could be a useful tool for stroke analysis. Moreover, the relatively unchanged, and high correlation of ODI and RDI maps following volume scaling, suggest that they may be sensitive to microstructure changes following stroke not present in the other models. While volume itself is not a predictor of stroke severity14, volume scaling the sKLD had a notable effect. The approach may lead to a more sophisticated measure that takes into account differences between brain hemispheres and is sensitive to volume depending on lesion location.
This study will continue to enroll subjects and validate the results obtained thus far. Time point analysis will be done to determine whether DSI models such as NODDI better capture ipsilesional and contralesional differences, and whether these differences show trends that could be used in clinical prognosis for ischemic stroke subjects.
[1] Kuo et al., Neuroimage, 41:7-18, 2008. [2] https://www.cmrr.umn.edu/multiband/ [3] Sullivan et al., Stroke 42:427-432, 2011 [4] http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/ [5] Zhang et al., Neuroimage, 61:1000-16, 2012 [6] http://mig.cs.ucl.ac.uk/index.php?n=Tutorial.NODDImatlab [7] Daducci et al., Neuroimage, 105:32-44, 2015 [8] https://github.com/daducci/AMICO [9] Adluru et al., IEEE EMBC, 36:742-745, 2014 [10] https://cran.r-project.org/web/packages/entropy/index.html [11] Gauthier et al., Stroke. 39(5): 1520–1525, 2008. [12] Granziera et al., Int Soc Magn Reson Med, p. 4199, 2011. [13] Granziera et al., Neurology, vol. 79, pp. 39-46, 2012. [14] Puig, American Journal of Neuroradiology, vol. 32, no. 5, pp. 857–863, Jul. 2011.