Quantitative diffusion tensor tractography of the motor and sensory tract in children with cerebral palsy

Authors


Dr Shoko Yoshida at Department of Radiology, Kyoto City Hospital, 1–2 Higashi Takada-cho, Mibu, Nakagyo-ku, Kyoto 604–8845, Japan. E-mail: sho-ko@mx.biwa.ne.jp

Abstract

Aim  The aim of this study was to compare the findings of quantitative diffusion tensor tractography of the motor and sensory tracts in children with cerebral palsy (CP) and typically developed comparison individuals, and also to evaluate the correlation with gross motor function.

Method  Thirty-four children with CP (mean age 2y 2.mo, SD 2y 0mo; 19 with spastic diplegia, eight with hemiplegia, six with spastic quadriplegia, and one with spastic triplegia) and 21 healthy comparison children (mean 2y 1.68mo, SD 2y 8.64mo) were evaluated. The distribution of Gross Motor Function Classification System (GMFCS) levels in the CP group was as follows: level I, 7; level II, 14; level III, 5; level IV, 3; and level V, 5. The following three diffusion tensor imaging (DTI) parameters including tractography were evaluated for each tract (corticospinal tract [CST] and posterior thalamic radiation [PTR]): number of fibres, tract-based fractional anisotropy, and region of interest (ROI)-based fractional anisotropy. We compared each value between the two groups, and correlated each value with the GMFCS level.

Results  The number of fibres and ROI-based fractional anisotropy values of both tracts were significantly lower in children with CP than in the comparison group (p<0.05–0.001). Additionally, there was significant negative correlation between GMFCS level and motor–sensory parameters (p<0.001–0.05).

Interpretation  DTI parameters of the CST and PTR in children with CP were significantly lower than in comparison children. In addition, these parameters were significantly correlated with GMFCS level.

List of Abbreviations
CST

Corticospinal tract

DTI

Diffusion tensor imaging

ROI

Region of interest

PTR

Posterior thalamic radiation

PVL

Periventricular leukomalacia

Cerebral palsy (CP) is defined as a group of non-progressive motor disorders of movement and posture due to a defect or lesion of the developing brain.1 Early accurate evaluation of individuals with, or at risk of, CP is critical because certain physiotherapy programmes can be initiated, and these offer the possibility of improving the neurological outcome.2 Despite a wide range of medical interventions in children with CP and at risk of CP, there is significant variability in outcome,3,4 in part related to the heterogeneous nature of the underlying brain pathology. Abnormalities are demonstrable by conventional magnetic resonance imaging (MRI) in many children with CP, but some children with typical or nearly typical MRI findings can have permanent motor dysfunction.5–7

Diffusion tensor imaging (DTI) has been proven useful for evaluating brain development and white matter injury,8 and some quantitative studies in children with, or at risk of, CP have examined the correlation between DTI data and clinical outcome.9–15 Recent studies using DTI have shown that the sensory pathways may be affected more severely than motor pathways,11 although injury to the corticospinal tract (CST) is thought to be the major determinant of motor impairment. On the other hand, some studies have reported that parameters related to the CST are significantly correlated with motor function outcome in children with periventricular leukomalacia (PVL).14 However, there have been few quantitative studies of both the motor and sensory tracts in children with CP, and it remains unclear if there is a correlation between tract defects and clinical prognosis.

Thus, in the present study, we compared the findings of quantitative diffusion tensor tractography of the CST and posterior thalamic radiation (PTR) in children with CP and typically developing comparison children, and we evaluated the correlation with gross motor function.

Method

Participants

We evaluated DTI performed between December 2006 and December 2008 in 34 children with CP (20 males, 14 females; mean age 2y 2mo, SD 2y 0mo, range 5mo–8y). The corrected age distribution of the children with CP was as follows: 0 to 6 months, n=1 (5mo); 6 to 11 months, n=7; 1 year, n=7; 2 years, n=6; 3 to 4 years, n=8; and 5 to 8 years (n=5). The mean gestational age of the group of children with CP was 34.2 weeks (SD 4.3wk; range 25–40wk), and the mean gestational weight was 2210g (SD 744g; range 774–3736g). The group comprised 19 preterm children and 15 term children. All participants were referred to our institutions because of the possibility of developmental delay. Neurological assessment of children with CP was performed by one of four paediatric neurologists (TK, YY, NY, or HH), each with at least 10 years’ paediatric neurological experience. Their assessment was based on Vojta’s motor risk criteria, which include spontaneous movements, postural reactions, and primitive reflexes.16 The neurologists were blind to the MRI findings of the children. The types of CP in the 34 children and the findings of conventional MRI were spastic diplegia in 19 (PVL, germinal matrix haemorrhage, diminished myelinated white matter, and symmetrical central tegmental tract hyperintensity), hemiplegia in eight (porencephaly, germinal matrix haemorrhage, unilateral polymicrogyria, schizencephaly, and cerebral infarction), spastic quadriplegia in six (cystic encephalomalacia, parasagittal cerebral injury, Rolandic cerebral palsy, and PVL), and spastic triplegia in one (porencephaly). Individuals with predominant athetosis and ataxic-type CP were excluded from the study. Participants’ Gross Motor Function Classification System (GMFCS) level was assessed at the same time as the MRI examination by the same four paediatric neurologists.

The comparison group included 21 typically developing children (nine males, 12 females; mean age 2y 2mo, SD 2y 9mo, range 4mo–9y) who were assessed to be developmentally typical by a paediatric neurologist (SO) blinded to the MRI findings. The corrected age distribution of the comparison children was as follows: 0 to 6 months, n=2 (4 and 5mo); 6 to 11 months, n=5; 1 year, n=3; 2 years, n=3; 3 to 4 years, n=4; and 5 to 8 years, n=3. The mean gestational age of the comparison group was 38.6 weeks (SD 1.7wk, range 35–41wk), and the mean gestational weight was 3016g (SD 427g, range 2140–3804g). Only one child was preterm. The reasons for MRI examination were benign seizure without neurological sequelae (n=10), epilepsy (n=5), examination for café au lait spots without neurological abnormality (n=2), nystagmus (n=1), headache (n=1), large head circumference (n=1), and setting sun sign (n=1).

This study was approved by the institutional review boards of the authors’ institutions, and written informed consent was obtained from the parents of each child.

MRI protocols

MRI was performed using a 1.5T scanner (Symphony; Siemens, Erlangen, Germany). All children aged 5 years or younger and children aged 6 years or older who could not remain still were sedated during MRI examinations using trichlorethyl phosphate syrup or chloral hydrate suppository under standard protocol use. Initially, all individuals underwent routine clinical pulse sequences, including sagittal and axial spin-echo T1-weighted imaging, axial and coronal fast spin-echo T2-weighted imaging, axial isotropic diffusion-weighted imaging using a single-shot, spin echo-type echoplanar sequence along three independent axes (b value=0, 500, 1000s/mm2, slice thickness 5mm), and T2*-weighted imaging. In all sequences, the field of view (FOV) was 200×200mm (number of images averaged [NEX]=1).

DTI was undertaken using a scheme of 12 different gradient directions along with four reference images (retention time [TR]/echo time [TE]=5700–5900/102). The maximum b value was 1000s/mm2. Spin echo acquisition and sensitivity encoding was used with an eight-channel head coil. The slice thickness was 3mm with no skip, matrix was 128×128, and the FOV was 200×200mm.

Postprocessing and fibre-tracking method

All DTI acquisition data sets were transferred offline. Postprocessing was performed using DTIstudio software (available at http://cmrm.med.jhmi.edu) and included generation of fractional anisotropy, vector maps, and colour-coded maps.

Fibre tracking was performed on the colour-coded map using the two-ROI method employing DTIstudio software.12,17–19 We drew the CST for evaluation of the motor tract and the PTR for evaluation of the sensory tract. To illustrate the CST, the first ROI was drawn in the posterior limb of the internal capsule and the second ROI was defined as the precentral gyrus, using an ‘AND’ operation. The most ventral axial slice that can clearly identify the cleavage of the central sulcus in the tracking result was selected and the bundle in the primary motor cortex was defined. As long as only the trajectories to the primary motor cortex are defined, the size of the second ROI can be arbitrary. The trajectories outside the two ROIs may cross the midline via the pontine crossing fibres and re-enter the contralateral hemisphere, which interferes with subsequent quantification procedures. These tracts should be removed by using the ‘NOT’ operation across the entire midsagittal slice. To illustrate the PTR, the first ROI was the retrolenticular part of the internal capsule and the second ROI was the thalamus (Fig. 1). It can be identified as a pair of green fibre bundles lateral to the corpus callosum. For the PTR reconstruction, fibres that were apparently unrelated to the tracts of interest, such as the corpus callosum and the anterior limb of the internal capsule, were rejected by using a ‘NOT’ operation.12 We confirmed the position of ROIs using both coronal and sagittal slices, as well as T2-weighted and T1-weighted imaging. The threshold chosen for fractional anisotropy was 0.15 and the angle threshold 60 degrees. These thresholds are lower than those used in adults because the fractional anisotropy of the white matter is lower in the paediatric brain than in the mature adult brain.12,14,15 To maintain consistency of the ROI placement, fibre tracking was performed by a single operator (SY), who was blinded to the children’s identity.

Figure 1.

 Schematic representation of drawing of tractography of the corticospinal tract (CST) and posterior thalamic radiation (PTR) using the colour-coded map. Left panel: To illustrate CST, we drew one region of interest on the posterior limb of the internal capsule (the blue fibres lateral to the thalamus at the level of globus pallidus), and the second region of interest defined the precentral gyrus at the level of high centrum semiovale. An ‘AND’ operation selected the fibres passing through both regions of interest. Right panel: To illustrate PTR, the first region of interest was drawn in the retrolenticular part of the internal capsule in the coronal plane at the level of the posterior cingulum and in the axial plane (the green fibre bundles lateral to the corpus callosum), and the second region of interest defined the thalamus.

Tracts and fractional anisotropy measurements

After fibre tracking, the following three measurements were performed. First, we measured the number of fibres forming the CST and PTR in the two groups. This was performed in a semi-automatic manner using the software (‘fibre statistics’). This is the parameter that demonstrates the volume of the depicted tract with fractional anisotropy above threshold in the brain parenchyma. Next, we measured the fractional anisotropy of each tract using two different methods. One method was measurement of the fractional anisotropy for each whole tract. This is the fractional anisotropy of all voxels that constitute the depicted tract, which was calculated in a semi-automatic manner (‘fibre statistics’). We called this method ‘tract-based fractional anisotropy (FA) measurement’. The other method involved measuring the fractional anisotropy of ROIs of the CST and PTR, which were drawn manually. The ROIs were set at the posterior limb of the internal capsule for the motor tract and the PTR for the sensory tract. The ROIs were drawn to follow the contour of the identified tract on the axial plane. We called this method ‘ROI-based fractional anisotropy (FA) measurement’. All ROIs were set by the same operator (SY) who performed fibre tracking. All measurements were performed twice, and the mean value was used for analysis.

Quantitative data analysis

These measurements provided two points for analysis. First, we compared each CST and PTR between the CP group and the comparison group. In the case of individuals with hemiplegia, we used only the measurements from the affected side, based on published findings that DTI measurements on the unaffected side are normal.13 The CP group included eight children with hemiplegia (four with right-sided hemiplegia and four with left-sided hemiplegia). We compared the parameters of each side (right and left) between the CP group (n=30) and the comparison group (n=21) using repeated-measures analysis of variance (ANOVA) and Welch’s t-test. Secondly, to evaluate their correlation with clinical condition, we correlated each measurement with the GMFCS level at the same time as MRI examination. GMFCS level was evaluated by four paediatric neurologists who were blinded to the MRI findings of the children. Children below 2 years of age were considered at their corrected age if they were born preterm. The distribution of GMFCS level in children with CP was as follows: level I, 7; level II, 14; level III, 2; level IV, 3; and level V, 5.

Statistical analysis

Statistical analysis was performed using Microsoft Excel 2007 for Windows, (Microsoft Corporation, Redmond, WA, USA), SPSS v.16 (SPSS Inc., Chicago, IL, USA), and PASW Statistics v.17.0. (IBM, NY, USA). Data are given as means and 2SD. Repeated-measures ANOVA using a mixed model and Welch’s t-test were used to analyse observed differences at a significance level of p<0.05. Spearman’s correlation coefficients were calculated to analyse associations. Intrarater reliability was assessed by using the interclass correlation coefficient (ICC) with 95% confidence intervals.

Results

Comparison of CP group and comparison group

There was no significant interaction between the right side and the left side in all parameters in either the CP group or the comparison group. Results for the various parameters in the CP group and the comparison group are shown in Table I (CST) and Table II (PTR). The number of fibres in the CST and ROI-based fractional anisotropy values were significantly lower in children with CP than in the comparison group (p<0.01). In the case of the PTR, the number of fibres (p<0.01), ROI-based fractional anisotropy values (left side, p<0.01; right side, p<0.05), and left-sided tract-based fractional anisotropy values (p<0.05) were significantly lower in children with CP than in the comparison group. Scatterplots showing, for each age, the number of fibres and ROI-based fractional anisotropy values in the two groups are shown in Fig. S1 (supporting information published online) for CST and in Fig. S2 (supporting information published online) for PTR. Intrarater reliability was assessed by using the ICC, and the ICC of each measurement ranged from 0.94 to 0.99.

Table I.   Comparison of the corticospinal (motor) tract in children with cerebral palsy and comparison children: number of fibres, tract-based fractional anisotropy (FA), and region-of-interest-based FA
 Number of fibresTract-based FAROI-based FA (PLIC)
RightLeftRightLeftRightLeft
MeanSDMeanSDMeanSDMeanSDMeanSDMeanSD
  1. ap<0.01 (Welch’s t-test). CP, cerebral palsy; ROI, region of interest; PLIC, posterior limb of internal capsule.

CP (right, n=30; left, n=30)3232633212340.500.060.500.120.510.520.550.14
Comparison (right, n=21;, left, n=21)6753036872600.520.040.540.040.630.060.650.07
p value<0.00a<0.00a0.1740.123<0.001b0.001b
Table II.   Comparison of the posterior thalamic radiation (PTR; sensory) in children with cerebral palsy and comparison children: number of fibres, tract-based fractional anisotropy (FA), and region-of interest-based FA
 Number of fibresTract-based FAROI-based fractional anisotropy (PTR)
RightLeftRightLeftRightLeft
MeanSDMeanSDMeanSDMeanSDMeanSDMeanSD
  1. ap<0.05, bp<0.01 (Welch’s t-test). CP, cerebral palsy; ROI, region of interest.

CP (right, n=30; left, n=30)1992011491070.460.070.470.080.430.120.430.12
Comparison (right, n=21; left, n=21)5842455392350.490.040.510.040.490.060.520.07
p value<0.001b<0.00b0.1580.027a0.022a0.001b

Correlation with clinical condition

Spearman’s correlation coefficient between GMFCS level and each tract is shown in Table III. There were good to strong significant negative correlations between the GMFCS and the parameters of the CST (i.e. the number of fibres [r=−0.632, −0.583] and ROI-based fractional anisotropy values [r=−0.616, −0.505]). In contrast, the correlations between the GMFCS and the parameters of the PTR were significantly strong to fair (i.e. the number of fibres [r=−0.623, −0.643] and ROI-based fractional anisotropy values [r=–0.366, –0.406]). There was no significant correlation between GMFCS level and tract-based fractional anisotropy values of the motor or sensory tract. To confirm that the parameters of interest were significantly correlated with GMFCS level, and not attributable to age differences, we added multiple regression analysis between GMFCS level, age, and motor–sensory parameters (Table IV).

Table III.   Correlation between GMFCS level and motor–sensory tract parameters (Spearman’s correlation coefficient)
 Correlation coefficient (r)
Motor (CST)Sensory (PTR)
RightLeftRightLeft
  1. ap<0.05, bp<0.01. CST, corticospinal tract; GMFCS, Gross Motor Function Classification System; PTR, posterior thalamic radiation; FA, fractional anisotropy; ROI, region of interest.

Number of fibresr−0.632−0.583−0.623−0.643
p<0.001b<0.001b<0.001b<0.001b
Tract-based FAr−0.290−0.102−0.237−0.276
p0.039a0.4770.0950.050
ROI-based FAr−0.616−0.505−0.366−0.402
p<0.001b<0.001b0.008b0.003b
Table IV.   Multiple regression analysis between GMFCS, age, and motor/sensory tract parameters
 Multiple coefficient of determination (r2)p value
Age (y)Parameters of CST/PTR
  1. ap<0.05, bp<0.01 (unpaired t-test). CST, corticospinal tract; GMFCS, Gross Motor Function Classification System; PTR, posterior thalamic radiation; FA, fractional anisotropy; ROI, region of interest.

Number of fibres (CST)Right0.3520.149<0.001b
Left0.389<0.0001b<0.00b
Tract-based FA (CST)Right0.3070.0580.001b
Left0.1160.4740.031a
ROI-based FA (CST)Right0.6330.007b<0.001b
Left0.5150.066<0.001b
Number of fibres (PTR)Right0.2290.4170.002b
Left0.2690.4710.005b
Tract-based FA (PTR)Right0.1650.2750.009b
Left0.2020.1670.003b
ROI-based FA (PTR)Right0.2750.0750.0004b
Left0.3660.046<0.001b

Discussion

In this study, our data demonstrated that the number of fibres and ROI-based fractional anisotropy comprising the CST and PTR were significantly lower in children with CP than in the comparison children. To the best of our knowledge, this has not been clearly reported by any previous study, although it is consistent with the clinical observation that many children with CP have sensory disturbance in addition to motor impairment.20 With regard to hemiplegia, Thomas et al.10 reported that there was a significant reduction in the number of fibres constituting the CST and superior thalamic radiation. With regard to PVL, some studies using visualized evaluation with colour maps and tractography have reported that the fibres connecting to the sensory cortex are markedly reduced in comparison with those of the CST in some individuals with severe PVL, suggesting that the posterior thalamocortical fibres might be more affected.11 Fan et al.21 reported that fractional anisotropy of both the motor and sensory pathways was reduced in children with PVL. With conventional MRI, it has also been reported that, in PVL, both the motor and the sensory tracts may be disturbed.22 Ludeman et al.15 using quantitative DTI evaluation (fractional anisotropy and transverse diffusivity), reported that the CST is affected more in children with permanent motor dysfunction than in children with a good prognoses.

Our study also demonstrated that motor tract parameters (number of fibres and ROI-based fractional anisotropy) were strongly and significantly correlated with GMFCS level. Several other studies have also demonstrated a correlation between quantitative motor tract parameters and clinical motor outcome.9,14,15,23 In addition, we found a strong to fair correlation between GMFCS level and PTR parameters (number of fibres and ROI-based fractional anisotropy), suggesting an involvement of the sensory tract. Hoon et al.24 showed that there was significant correlation between the DTI score of the thalamocortical pathways and quantitative clinical status (including motor and sensory function) of children with CP; we suggest that motor and sensory function cannot develop independently, and that both motor and sensory pathways play a significant role in the development of clinical function in children with CP.

We measured the following three parameters in this study: number of fibres, tract-based fractional anisotropy, and ROI-based fractional anisotropy. We found that there was a discrepancy between the number of fibres/ROI-based fractional anisotropy findings and tract-based fractional anisotropy findings, with the former tending to be more strongly correlated with GMCFS level than the latter. We suggest that this discrepancy is probably due to the fact that, in the tract-based analysis, the threshold is based on fractional anisotropy. In fibre tracking using the two-ROI method, fibres below the threshold fractional anisotropy were ignored, and so the corresponding tracts were not included in the calculation of tract-based fractional anisotropy measurements. Thus, the difference is thought to be reflected in the number of fibres and not in the tract-based fractional anisotropy.

With regard to the DTI borderline between children with CP and typically developing children, Murakami et al.14 suggested that a tract-based fractional anisotropy value of <0.5 in the CST may be a useful threshold for identifying children with PVL with permanent motor dysfunction.14 Our results demonstrated that the distributions of each fractional anisotropy value overlapped (both tract-based fractional anisotropy and ROI-based fractional anisotropy) between the CP and comparison groups (Figs S1 and S2) for CST and PTR, and the fractional anisotropy threshold could not be determined. We were unable to evaluate the correlation between the parameters and long-term motor outcomes of children with CP; however, we suggest that the DTI parameters of the CST and PTR are useful variables for evaluation of clinical motor condition and outcome in children with CP.

There are several further limitations of our study. Intrarater agreement for both tract-based and ROI-based measurements was good, but we could not assess interrater reliability because fibre tracking, ROI drawing, and measurements were performed by the same person. Another limitation was that we included children with a wide age range. DTI values change rapidly in the first 2 years of life, especially in the first 6 months.25 In this study, we did not perform analysis for each age group, because the number of children in each group was small. However, it will be necessary to carry out a further study, including a large number of children with CP, divided into subgroups such as under 6 months of age, between 7 months and 2 years, and over 2 years. Additionally, we did not attempt to correlate clinical sensory function with the evaluated parameters. Many children with CP have sensory impairment,20 and its correlation with motor and sensory parameters may partly explain the pathogenesis, although objective evaluation of sensory function in children with CP is difficult.

In conclusion, quantitative evaluation of DTI was shown to be useful for assessment of children with motor dysfunction. Children with CP were found to have a significant reduction in CST and PTR. In addition, motor and sensory tract parameters were significantly correlated with GMFCS level.

What this paper adds

  •  Quantitative evaluation of DTI is shown to be useful for the assessment of children with motor dysfunction.
  •  CST and PTR parameters are significantly lower in children with CP than in typically developing children.
  •  CST and PTR parameters are significantly correlated with GMFCS level.

Ancillary