Long term motor function after neonatal stroke: Lesion localization above all

Abstract

Motor outcome is variable following neonatal arterial ischemic stroke (NAIS). We analyzed the relationship between lesion characteristics on brain MRI and motor function in children who had suffered from NAIS. Thirty eight full term born children with unilateral NAIS were investigated at the age of seven. 3D T1‐ and 3D FLAIR‐weighted MR images were acquired on a 3T MRI scanner. Lesion characteristics were compared between patients with and without cerebral palsy (CP) using the following approaches: lesion localization either using a category‐based analysis, lesion mapping as well as voxel‐based lesion‐symptom mapping (VLSM). Using diffusion‐weighted imaging the microstructure of the cortico‐spinal tract (CST) was related to the status of CP by measuring DTI parameters. Whereas children with lesions sparing the primary motor system did not develop CP, CP was always present when extensive lesions damaged at least two brain structures involving the motor system. The VLSM approach provided a statistical map that confirmed the cortical lesions in the primary motor system and revealed that CP was highly correlated with lesions in close proximity to the CST. In children with CP, diffusion parameters indicated microstructural changes in the CST at the level of internal capsule and the centrum semiovale. White matter damage of the CST in centrum semiovale was a highly reproducible marker of CP. This is the first description of the implication of this latter region in motor impairment after NAIS. In conclusion, CP in childhood was closely linked to the location of the infarct in the motor system. Hum Brain Mapp 36:4793–4807, 2015.. © 2015 Wiley Periodicals, Inc.

Abbreviations

AD

axial diffusitvity

BG

basal ganglia

BFMF

Bimanual Fine Motor Function

NAIS

Neonatal arterial ischemic stroke

CP

cerebral palsy

CSO

centrum semiovale

CST

cortico‐spinal tract

DTI

Diffusion tensor imaging

DWI

diffusion weighted imaging

FA

Fractional anisotropy

FDR

false discovery rate

GMFCS

Gross Motor Function Classification System

MCA

middle cerebral artery

MD

mean diffusivity

NAIS

neonatal arterial ischemic stroke

NPM

non‐parametric mapping software

RD

radial diffusivity

VLSM

voxel‐based lesion‐symptom mapping

INTRODUCTION

With a prevalence of 1/2,000 to 1/4,000 live births [Chabrier et al., 2011; Kirton and Deveber, 2013], perinatal ischemic stroke is the most frequent form of childhood stroke and constitutes the leading cause of unilateral cerebral palsy (CP) in term‐born children [Kirton and Deveber, 2013]. Perinatal ischemic stroke is an umbrella term including several conditions that differ in pathophysiology, timing and thus in outcomes [Kirton and Deveber, 2013]. Neonatal arterial ischemic stroke (NAIS) refers to a perinatal ischemic stroke syndrome with neonatal signs (mainly iterative focal seizures in the first days of life) related to an arterial infarct as revealed by brain imaging [Raju et al., 2007]. A more recent approach coins this term to neonates with MRI (notably through the use of diffusion‐weighted sequences) demonstrating the acute appearance of the arterial infarct [Kirton and Deveber, 2013]. NAIS differs from presumed perinatal ischemic stroke, in which clinical manifestations are delayed [Kirton et al., 2010; Rutherford et al., 2012].

NAIS leads to unilateral CP in approximatively 30% of cases [Chabrier et al., 2011]. According to the literature [Kirton and Deveber, 2013], the impact of NAIS on motor recovery and long‐term outcome depends on infarct characteristics in the neonatal period including location and lesion size. Typically, extensive damage to the basal ganglia (BG), the cortex and the posterior limb of the internal capsule is constantly associated with CP [Boardman et al., 2005]. Acute diffusion‐weighted MRI changes in the cortico‐spinal tract (CST) correlate with poor motor outcome [De Vries et al., 2005; Domi et al., 2009; Kirton et al., 2007]. Inversely, the absence of involvement of the CST on early MRI (i.e., <28 days of life) as judged from visual inspection resulted in normal motor development at two years of age in 94% of cases within our cohort [Husson et al., 2010].

Whereas early infarct volume and location may help predict long‐term motor outcome especially in large lesions [Boardman et al., 2005; Lee et al., 2005; Mercuri et al., 1999; Mercuri et al., 2004], the prediction of motor outcome for smaller infarcts remains challenging on initial imaging [Hashimoto et al., 2001; Kirton et al., 2007]. For example, limited CST involvement could evolve favourably [Husson et al., 2010]. In case of mild motor impairment or moderate CP, infarct location and lesion size at the acute stage showed limited correlation with long‐term motor outcome [Mercuri et al., 1999; Mercuri et al., 2004]. In fact, most neuroanatomical correlates of motor deficit were derived from series including a limited number of patients with large age ranges at the time of the motor evaluation, or mixing patients with arterial and/or venous infarctions, or with both NAIS and presumed perinatal ischemic strokes, whose motor outcomes differ.

Obviously, age at motor evaluation appears crucial when studying the relation between lesion characteristics and motor function, as new critical fine motor skills develop until late childhood, and thus the full consequences of the lesion on motor function evolve with late brain maturation. Moreover, the early changes of imaging characteristics following NAIS may complicate the estimation of motor prognosis in acute settings. Indeed, diffusion‐weighted imaging may miss or underestimate the extent of brain lesions in the first 2 days after injury [McKinstry et al., 2002]. Follow‐up MRIs after the NAIS [de Vries et al., 1997] does allow better delimitation of cerebral lesions and help refine the study of neuroanatomical correlates of motor function following NAIS.

Little is known about the anatomo‐clinical correlations of motor deficit in school‐age children having suffered from NAIS when both clinical status and lesions are stabilized. Therefore, the aim of this study was to analyze the relationship between brain lesions and motor function in 7 year‐old children who had suffered from NAIS in the territory of the middle cerebral artery (MCA).

METHODS

Participants and Neurological Examination

Patients belonged to the previously described AVCnn cohort (Accident Vasculaire Cérébral du nouveau‐né, that is, neonatal stroke—PHRC regional n°0308052 and PHRC interregional n°1008026 ‐eudract number 2010‐A00329‐30) [Chabrier et al., 2010; Husson et al., 2010]. In brief, 100 term newborns with an arterial infarct, confirmed by brain imaging (CT and/or MRI), who were symptomatic in the neonatal period (thus matching the 2007 definition of NAIS [Raju et al., 2007]) were consecutively enrolled between November 2003 and October 2006 from 39 neonatology and neuropediatrics units distributed throughout mainland France. At the age of 7 years, a clinical follow‐up visit was organized and an MRI investigation was proposed to the 80 children who took part in the clinical evaluation. Of the 52 children who participated in the MRI study, 38 had unilateral MRI lesions in the MCA territory. They constituted the population of this study (eudract number 2010‐A00976‐33).

Informed written consent from patients/parents and approval from the ethical committee of the university hospital of Angers, France, were obtained.

Neurological examination was performed by paediatric neurologists (MK, SC, SN) or a paediatric physical and rehabilitation medicine practitioner (MD) to assess the presence or absence of unilateral CP [Bax et al., 2005; Rosenbaum et al., 2007] according to the SCPE‐CP‐decision‐tree (SCPE, 2000).

For detailed characterization of CP patients, motor function was classified according to the Gross Motor Function Classification System (GMFCS) [Palisano et al., 1997; Palisano et al., 2008]. The GMFCS is a validated five level classification system to stratify CP patients according to their level of independent gross motor function on the basis of self‐initiated movements with particular emphasis on sitting, walking, and wheeled mobility. The levels could be summarized as: Level I: subject walks without limitations, II: walks with limitations, III: walks using a hand‐held mobility device (crutches or canes…), IV: self‐mobility with limitations may use powered mobility, and V: transported in a manual wheelchair.

Handedness was assessed according to the Edinburgh inventory [Oldfield, 1971]. Hand function was assessed by the Bimanual Fine Motor Function (BFMF) [Beckung and Hagberg, 2002], which classifies fine motor function into five levels, with Level I indicating the least severe and Level V the most severe limitation, to assess hand function in children with CP [Christine et al., 2007].

Structural MRI

Acquisition

Images were acquired on a 3.0 Tesla scanner (MAGNETOM Trio Tim system, Siemens, Erlangen, Germany, 12 channel head coil) at Neurospin, CEA‐Saclay, France.

Imaging sequences included a high‐resolution 3D T1‐weighted volume using a magnetization‐prepared rapid acquisition gradient‐echo sequence [176 slices, repetition time (TR) 2300 msec, echo time (TE) 4.18 msec, field of view (FOV) 256 mm, flip angle = 9°, voxel size 1 × 1 × 1 mm³], a 3D FLAIR sequence [160 slices, TR 5,000 msec, TE 395msec, FOV 230 mm, voxel size 0.9 × 0.9 × 1 mm³], and a diffusion‐weighted dual SE‐EPI sequence with 30 diffusion encoding directions and a diffusion‐weighting of b = 1,000 s/mm² (TR = 9,500 msec, TE= 86 msec, 40 slices, voxel size 1.875 × 1.875 × 3 mm).

MRI analysis and classification of the MCA infarct location

3D T1‐weighted and 3D FLAIR images of each participant were assessed visually by consensus (two senior paediatric neuroradiologists, LHP, BH, and four physical and rehabilitation medicine practitioner or paediatric neurologists, MD, MK, SC, SN) to visually identify and characterize the lesions.

In order to perform a category‐based analysis, lesions were classified depending on the damaged brain structures as follows [Boardman et al., 2005]: 1—lesions in the primary motor system, that is, precentral gyrus, basal ganglia (BG)/thalamus, internal capsule; 2—lesions outside the primary motor system [i.e., frontal (except the precentral gyrus), parietal, occipital, temporal lobes and insula].

Lesion delineation and normalization

For each patient, the boundaries of the lesion were manually delineated on a slice by slice basis by two of the authors (MD, SG) on the individual 3D T1 images to create a binary lesion mask using the MRIcron software (http://www.mccauslandcenter.sc.edu/mricro) [Rorden et al., 2007]. In case of a main branch MCA stroke, the lateral border of the lesion mask was drawn along the inner border of the skull, comprising the whole porencephaly. The co‐registered 3D FLAIR images were used as a visual aid for precise lesion location. The hyperintense areas on FLAIR (“gliosis”) were defined as part of the lesion. The two raters were blinded for clinical information, especially motor function. Initially they separately delineated the lesions, later they were double‐checked and validated by consensus, then projected on to the left hemisphere by flipping the image in the left‐right direction. This was achieved by modifying the orientation matrix in the header, which flips the image in the left/right [X‐] dimension.

T1‐weighted 3D volume data and lesion masks from all subjects were spatially normalized to a customized symmetric paediatric brain template using the unified segmentation‐normalization algorithm [Ashburner and Friston, 2005] of the statistical parametrical mapping software, SPM8 (Wellcome Department of Imaging Neuroscience, University College, London, UK; http://www.fil.ion.ucl.ac.uk/spm) running in Matlab R2011a (The MathWorks, Natick, MA). Lesion masks were used to mask out abnormal tissue during the spatial normalization routine [Wilke et al., 2011]. The customized paediatric template was created using the Template‐O‐Matic Toolbox (http://irc.cchmc.org/software/tom.php, [Wilke et al., 2008]). After spatial normalization of the T1‐weighted 3D image to the template space, the resulting spatial transformation field was applied to the lesion masks in order to warp the binary lesion masks into the common normalized (MNI) template space with a resolution of 1 × 1 × 1 mm.

To avoid bias due to variable brain sizes, a lesion volume ratio was calculated as follow: total intracranial volume (TIV) was approximated for each participant by calculating the sum of gray matter, white matter and cerebro spinal fluid maps obtained from the preprocessing steps, then a lesion volume ratio for each participant was calculated as lesion Volume Ratio = (lesion volume/TIV) *100. These lesion volume ratios were compared between CP and non CP patients using the non‐parametrical Mann‐Whitney U test. Significance was assumed at P < 0.05.

Lesion maps

Group lesion maps were created by summing up all individual normalized lesion maps, to identify damaged regions common to multiple patients. This was done separately for the CP (Fig. 2A) and non CP (Fig. 2B) groups. Group lesion maps in CP and non‐CP patients were overlaid on an MNI template brain with a display threshold of n > 4 [Lo et al., 2010]. Exact MNI coordinates of the highest overlaps were obtained using the MRIcron software (http://www.mccauslandcenter.sc.edu/mricro/).

Figure 2.

Lesion overlap plots for the group of: A, patients with unilateral CP (n = 14); B, patients without CP (n = 24). The colour range indicates the number of overlapping lesions by coding increasing frequencies from violet (n = 1 subject) to red (n = maximum number of subjects in the respective group). [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

VLSM analysis

We used a voxel‐based lesion‐symptom mapping (VLSM) approach to calculate a test statistic at each voxel correlating lesions with motor outcome (CP or not) [Bates et al., 2003]. VLSM provides a robust alternative to lesion‐characterising studies [Rorden and Karnath, 2004] traditionally used to understand which brain areas (location or spatial extent) are critical to preserve motor function in chronic stroke patients [Lo et al., 2010; Schoch et al., 2006].

For statistical analysis, all voxels in which at least 5% of the patients had a lesion (i.e., 2/38) were included in the VLSM analysis [Geva et al., 2012]. The presence or absence of unilateral CP was first tested (binomial test) using the non‐parametric mapping (NPM) software [Rorden et al., 2007]. The resulting statistical measure was used to create a map identifying brain areas with significant correlation of lesion likelihood with neuromotor status (absence or presence of CP), after correction for multiple comparisons using the false discovery rate (FDR) (P < 0.05). This correction for multiple comparisons was modified by employing the non‐parametric permutation test recommended for medium‐sized samples [Medina et al., 2010]. In addition, a VLSM analysis was conducted using lesion size as covariate of non interest [Kimberg et al., 2007].

Analysis of CST microstructure

To explore the presumed functionally relevant white matter part of the motor system, the microstructure of the corticospinal tract was analyzed. Diffusion‐weighted imaging data was available in 22 children without CP and 11 with CP. Patient's 1, 22, 33, 37, and 38 (in Table 1) had no diffusion‐weighted data available because of artefacts due to head movements. As described before [Groeschel et al., 2014], the CST was reconstructed using probabilistic tractography based on constrained spherical deconvolution [Tournier et al., 2007]. To quantify CST microstructure, diffusion parameters (fractional anisotropy (FA), axial (AD), radial diffusivity (RD) and mean diffusivity (MD)) were measured in the CST at the level of the internal capsule and the centrum semiovale (CSO) in all subjects (see Fig. 4), as previously done [Groeschel et al., 2014].

Table 1.

Details of clinical and MRI findings

	Clinical description					MRI findings
No.	Sex	Motor Examination	Handedness	GMFCS	BFMF	Lesion side	Precentral gyrus involvement	Basal Ganglia/thalamus involvement	Internal capsule involvement	Lesioned non motor hemisphere
1	F	N	R			L	–	–	–	P, I
2	M	L hemi	R	I	II	R	Y	–	–	Fr, P, T, I
3	M	N	L			L	–	–	–	P, O, I
4	M	N	R			R	–	–	–	P
5	M	N	L			L	–	–	–	P, I
6	F	N	R			L	–	Y	–	–
7	F	N	R			R	–	–	–	P, I
8	M	L hemi	R	I	I	R	Y	–	–	Fr, P, I
9	F	N	R			L	Y	–	–	Fr, P, T, I
10	F	L hemi	R	I	I	R	Y	Y	–	Fr, P, T, I
11	M	N	R			L	–	–	–	P
12	F	L hemi	R	III	I	R	Y	Y	Y	Fr, P, T, I
13	M	N	L			L	–	–	–	P
14	F	N	R			R	–	–	–	P, I
15	M	N	R			L	Y	–	–	Fr, P, I
16	F	N	L			L	–	Y	–	–
17	M	N	R			R	–	–	–	P
18	F	N	L			L	–	–	–	P, T, I
19	M	N	R			L	Y	–	–	Fr, P
20	M	N	R			R	–	–	–	P, I
21	F	N	L			L	Y	–	–	Fr, P, I
22	F	R hemi	L	I	II	L	Y	Y	–	Fr, P, T, I
23	M	N	R			R	–	–	–	P
24	M	R hemi	L	II	II	L	Y	Y	Y	Fr, P, T, I
25	M	N	L			L	–	Y	–	T, I
26	M	R hemi	L	II	I	L	–	Y	–	P, T, I
27	F	N	R			R	–	–	–	I
28	M	R hemi	L	I	I	L	–	Y	–	P, I
29	F	R hemi	L	III	I	L	–	Y	Y	P
30	M	N	L			R	–	–	–	P
31	F	N	R			L	–	–	–	P, T, I
32	F	R hemi	L	I	II	L	Y	Y	Y	Fr, P, T, I
33	M	R hemi	L	I	III	L	Y	Y	Y	Fr, P, T, I
34	M	R hemi	L	III	II	L	–	Y	Y	Fr, P, I
35	M	N	R			R	–	Y	–	–
36	M	L hemi	R	I	II	R	Y	–	–	Fr, I
37	M	L hemi	R	I	III	R	–	Y	Y	P, T, I
38	M	N	R			L	–	–	–	P

F, female; M, male; R, right; L, left; N: Normal, hemi, hemiplegia; BFMF, Bimanual Fine Motor Function; GMFCS, Gross Motor Function Classification System, ‐, no; Y, yes; Fr, frontal; P, parietal; T, temporal; O, occipital; I, insula. . see text for details.

Figure 4.

Mean diffusivity (MD, in mm²/s) computed in the corticospinal tract (CST) of patients with CP (n = 11) and without CP (n = 22). Box plots show higher MD in CST ipsilateral to the lesion at the level of the internal capsule and centrum semiovale in stroke children with CP compared to those without CP (* = p < 0.01). Contralateral to the lesion there was no difference between the CP and the non‐CP group. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

We deliberately focused our hypotheses on these two CST regions because they are known to be crucial in order to maintain motor function following stroke in adults [Gauthier et al., 2009; Lindenberg et al., 2010; Lo et al., 2010] and they easily can be anatomically characterized. The main hypothesis was an increased MD in the CST both at the CSO as well as at the level of the internal capsule in the CP compared to the non‐CP group. Because of the difficult interpretation of FA, AD, and RD at the level of the CSO, due to complex crossing fiber architecture [Groeschel et al., 2014; Wheeler‐Kingshott and Cercignani, 2009], we focused our hypothesis on the internal capsule expecting lower FA and higher RD in the CP group. Parameters were measured separately for the ispilesional and contra‐lesional side and compared between the CP and non‐CP group using the Mann–Whitney U test.

Post‐hoc analysis

In a post‐hoc analysis, in order to explore the added value of DWI for this purpose, we excluded those subjects with visual lesions in the internal capsule and compared diffusion parameters between the CP and the non‐CP groups at this location using the Mann‐Whitney U test.

In order to help disentangle between direct ischemic insult and Wallerian degeneration of the CST involvement in the subcortical CSO, we visually assessed the central part of CSO in children who developed later CP. Therefore we analyzed retrospectively whether obvious abnormalities were already present in neonatal imaging. A visual (qualitative) assessment was chosen, as the neonatal imaging data in the 14 children with CP was very heterogeneous (e.g., 11 had MRI and DWI; 2 only MRI; and 1 only CT) and therefore not suitable for quantitative analysis [Husson et al., 2010].

RESULTS

Patient Characteristics and Neuromotor Examination

Details of neuromotor examination and patient characteristics are presented in Table 1. In our population (23 boys, 15 girls), 14 children (37%, nine boys) had unilateral CP, six on the left (43%) and eight (57%) on the right side. All of them were able to walk. Their GFMCS grades were as follows: Grade I: n = 9, Grade II: n = 2, Grade III: n = 3. Regarding the hand function, six children were BFMF Grade I, eight were graded II or III.

Lesion Category‐Based Analysis

The details of the radiological extent of the lesions on MRI are given in Table 1.

Regarding the infarct location, 16 children (42%) had hemispheric lesions outside the motor system and none of them had CP.

Twenty‐two children (58%) had a lesion within the primary motor system, 14 of them (64%) had CP. In this latter group, seven patients had at least one lesion in the internal capsule, two had an intact internal capsule but lesions involving both the precentral gyrus and the BG/thalamus, and the remaining five patients had a preserved internal capsule but a lesion either in the precentral gyrus (n = 2) or in BG/thalamus (n = 3). All these 14 patients also had lesion extension outside the primary motor system. In the remaining eight patients (8/22, 36%) with lesions within the motor system but without CP, no child had infarct located in the internal capsule, three had isolated BG/thalamus lesion (and no hemispheric lesion) and five had either precentral lesion (n = 4) or BG/thalamus lesion (n = 1) associated with lesion extension outside the primary motor system.

The non‐CP group (n = 24, 63%) included 16 children with lesions outside the motor system (66%), three patients with isolated BG/thalamus lesion and five children with either BG/thalamus or precentral gyrus lesions associated with a lesion outside the primary motor system.

These results are summed up in a classification tree (Fig. 1).

Figure 1.

Classification‐tree relating presence or absence of unilateral CP with the presence or absence of brain damage. Hemi: unilateral cerebral palsy; BG: basal ganglia; numbers in boxes represent number of subjects. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Comparison of Lesion Volume Ratio between CP And non‐CP Groups

Lesion volume ratio was significantly higher in the CP group than in the non‐CP group (respectively: median = 6.46, 90% confidence interval [4.08–14.08] for CP group; median = 0.96, 90% confidence interval [0.78–2.04], for non CP group, P < 0.0001).

Group Lesion Maps

In unilateral CP patients (Fig. 2A) lesions appeared more extensive and more distributed than in non CP subjects (Fig. 2B). Compared to the latter group, they involved more frequently the precentral gyrus, frontal areas and the rolandic operculum, the basal ganglia, and the central white matter encompassing the CST in the CSO and the internal capsule. Inversely, in the non‐CP group lesions were more focal, more posterior (superior and inferior parietal lobules, and parietal operculum) and always preserved the BG and the inferior frontal lobe. MNI coordinates of the lesion overlap maps characterizing the lesions regarding absence or presence of CP are presented in Table 2.

Table 2.

MNI coordinates of the lesion overlap maps characterizing the lesions from our population regarding the absence or presence of CP

	Patients with CP		Patients without CP
	Number of patients	MNI coordinates [x,y,z]	Number of patients	MNI coordinates [x, y, z]
Rolandic Operculum	12	−43, −13, 13	12	−39, −31, 20
Insula	12	−41, −13, 11	12	−34, −25, 20
Postcentral gyrus	11	−49, −17, 21	10	−37, −33, 42
Inferior parietal gyrus	11	−51, −25, 37	10	−49, −31, 42
Supramarginal gyrus	11	−47, −27, 25	9	−44, −34, 24
Heschl'sg gyrus	11	−43, −13, 11	9	−53, −9, 10
Superior corona radiata	11	−32, −18, 20
Superior longitudinal fasciculus	11	−36, −12, 22	8	−38, −34, 26
Precentral gyrus	10	−49, −6, 22
Posterior corona radiata	10	−30, −24, 22	5	−30, −24, 22
External capsule	10	−34, −12, 8	5	−30, −22, 18
Superior temporal gyrus	9	−53, −9, −1	11	−43, −31, 17
Inferior frontal operculum	8	−48, 9, 4
Putamen	8	−33, −9, 5
Superior temporal pole	8	−57, 6, 0
Middle temporal gyrus	8	−51, 17, −2	6	−55, −31, 9
Inferior frontal gyrus	7	−59, 17, 5
Middle frontal gyrus	6	−35, 19, 31
Inferior orbitofrontal grus	6	−52, 20, −5
Angular gyrus	6	−55, −57, 26	5	−43, −49, 23
Pallidum	6	−25, −1, −2
Anterior limb of internal capsule	6	−22, 0, 12
Posterior limb of internal capsule	6	−26, −12, 10
Retrolenticular part of internal capsule	6	−32, −26, 8
Anterior corona radiata	6	−28, 26, 10
Sagittal stratum	5	−38, −24, −4
Caudate	5	−16, 17, 4

In bold characters regions implicated in the motor system

VLSM

The Figure 3 illustrates brain areas statistically correlated to unilateral CP. The VLSM analysis identified large regions overlapping cortical regions of the sensorimotor system and white matter encompassing the CST, along with the middle part of the putamen.

Figure 3.

VLSM identified lesioned regions most correlated with unilateral CP (p < 0.05, FDR correction). Voxels lesioned in at least 5% of patients were included. Colours represent z‐scores. Montreal Neurological Institute x, y, and z coordinates are given. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Areas showing the highest correlation with unilateral CP status were: the insula ([−33 −17 22], z = 4.76), the corona radiata ([−32 −20 22], z = 4.57), and notably the junction of the corona radiata and the corticospinal tract, the precentral gyrus ([−58 0 23], z = 4.50), the postcentral gyrus ([−61 −1 23], z = 4.50), the BG ([−32 2 8], z = 4.13), the rolandic operculum ([−61 4 2], z = 4.13), the external capsule ([−34 −8 8], z = 4.13) and the internal capsule (posterior limb) ([−26 −8 16], z = 3.20).

Only the involvement of superior longitudinal fasciculus, corona radiata (junction of these regions), insula, precentral and postcentral gyri survived at a more stringent correction for multiple comparisons (P < 0.05, FWE). Entering the lesion volume ratio as a covariate did not change the pattern of VLSM results, although the power of the analyses was consequently reduced.

The VLSM analysis thus showed that areas best correlated with the presence of unilateral CP were cortical structures involved in sensorimotor system (precentral gyri) but also white matter structures (CST) and particularly the junction of the superior longitudinal fasciculus, the corona radiata and the CST (centrum semiovale).

CST Microstructure

Children with CP had on average a higher MD in the ipsilesional CST than children without CP, both at the level of the CSO (median = 0.000856 mm²/s, 90% confidence interval [0.000787–0. 001030] versus 0.000763 mm²/s [0.000752–0.000811], P < 0.01) and at the level of the internal capsule (median 0.000835 mm²/s [0.000680–0.001187] versus 0.00755 mm²/s [0.000735–0.000760], P < 0.01). MD of the contralesional side did not differ between the groups (Fig.4).

In the internal capsule, we found a significant increase of RD on the ipsilesional CST in CP (median 0.000496, CI [0.00033–0.00092]) versus non‐CP (median 0.000408, CI [0.00038–0.00042], P < 0.001) without significant changes in AD (P = 0.154), resulting in a significant decrease of FA on the ipsilesional CST in CP (median 0.59, CI [0.45–0.65]) versus non‐CP (0.65, CI [0.64–0.67], P < 0.001). No differences were found on the contra‐lesional side.

In the CSO, we found again a significant increase of RD on the ipsilesional CST in CP (median 0.000682, CI [0.00058–0.00085]) versus non‐CP (median 0.000575, CI [0.00056–0.00062], P = 0.021) as well as of AD in CP (median 0.00123, CI [0.00118–0.00140]) versus non‐CP (median 0.00116, CI [0.00113–0.00120], P = 0.004), resulting in a non‐significant decrease of FA on the ipsilesional CST in CP (median 0.39, CI [0.33–0.45]) versus non‐CP (0.43, CI [0.41–0.45], P = 0.154). No difference were found on the contralesional side.

Results remained similar when excluding subjects with radiological lesions in the internal capsule (n = 5, Patients 12,24,29,32, and 34): MD was significantly increased in the CP group compared to the non‐CP group (P < 0.01), with increased RD (P < 0.001) and no change in AD (P = 0.304), resulting in a decreased FA (P < 0.001) in the CST at the level of the internal capsule.

An example of diffusion results is provided on Figure 5. A 3D FLAIR image of Subject 10 (who had unilateral CP with a precentral cortical lesion close to the CST and white matter gliosis surrounding the cortical lesion) is presented, with overlaid CST fibers. In addition, VSLM‐group‐results (FDR‐corrected at P < 0.05) were co‐registered and overlaid in order to demonstrate the close relationship between the cortical lesion and the involvement of the CST in the centrum semiovale, which appeared to be the most common lesion location amongst children with CP in this series.

Figure 5.

VLSM‐group‐results (FDR‐thresholded at p < 0.05) co‐registered and overlayed on 3D‐Flair coronal (top) and axial (bottom) slices of a single 7‐year old child with unilateral spastic CP (subject 10, see Table I). The CST of that child is also shown (in blue, estimated from probabilistic fiber tracking using diffusion‐weighted MRI to demonstrate the relationship between critical regions for developing CP and the cortico‐spinal tract. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Comparison with Neonatal Imaging

Of the 14 subjects who developed CP, all of the 11 subjects who had both neonatal MRI and DWI, had an obvious involvement of the central part of the CSO (for an example, see Fig. 6), consistent with a direct ischemic insult in the neonatal period. In one additional patient who only had a CT scan, CSO involvement was highly suspicious. In the two remaining CP patients with neonatal MRI but without DWI, there was no clear evidence for white matter involvement in the CSO, one of them having basal ganglia infarct and the other one a large cortical infarct of the central and postcentral regions.

Figure 6.

Imaging follow‐up of child with NAIS on the right hemisphere (subject 10 with unilateral (left) CP, see table I). A, DW Imaging in the neonatal acute phase: While cortical involvement is predominant, ischemic injury also involves subcortical white matter in the centrum semiovale at the location of the fanning of the corticospinal tract (compare to Fig. 5, same subject), B: T1‐weighted image and FLAIR‐image at 7 years of age of the same child: The atrophic lesion closely matches the pattern seen in the neonatal period, with phantom cortical ribbon in the central region, signs of gliosis of the subcortical and deep white matter, encompassing part of the cortico‐spinal tract (in red). [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

DISCUSSION

To our knowledge this is the first study characterising anatomo‐clinical correlations of brain lesions in children who were prospectively followed‐up from birth until school age after a NAIS. These results were obtained in a large homogeneous cohort in terms of patients' characteristics (term newborns, now aged seven), lesions' characteristics (precise timing of the injury in the neonatal period, infarct localization in the MCA territory) and imaging conditions (monocentric data acquisition using one 3 T MRI scanner). Therefore our results are not blurred by the developmental consequences of prematurity, the diversity in the acquisition of motor skills with age, or experimental multicentre acquisition variability.

Results confirm that the localization of the infarct into cortical and subcortical regions classically considered as part of the primary motor system like precentral gyri and CST largely explains the occurrence of CP. Children with lesions sparing the primary motor system did not develop CP. On the contrary, all children presenting with damage to the internal capsule, extensive lesions in motor cortical, subcortical or white matter structures developed unilateral CP. Our study corroborates previous studies using lesion category‐based analysis in the acute phase [Boardman et al., 2005; Mercuri et al., 2004]. These authors found that children with a combination of hemispheric, BG, and internal capsule involvement were at high risk of developing unilateral CP.

VLSM analysis enables us to study more accurately the relationship between motor outcome and lesion topography by reducing the influence of lesion volume, which is known to correlate with outcome. Indeed, a previous study [Ganesan et al., 1999] on ischaemic stroke in childhood has shown that in the case of MCA‐territory infarcts, lesion volume >10% of the intracranial volume was associated with poor motor outcome. In the current study, we were able to confirm that larger lesion volume is related to the occurrence of CP, but some children with small lesions also had poor motor outcome. Moreover, in adults with chronic stroke, no relationship was found between lesion volume and motor deficit [Mark et al., 2008; Page et al., 2013], neither was there in the acute phase [Saunders et al., 1995]. Clearly, further analysis is needed comparing lesion characteristics in acute and chronic phases of NAIS.

When comparing lesion location in CP and non‐CP NAIS children, we also confirmed that lesions in the CST are strongly related to motor outcome. Damage to the CST in the posterior limb of the internal capsule and/or cerebral peduncle was shown to accurately predict motor outcome in neonates [De Vries et al., 1999; Husson et al., 2010; Rutherford et al., 1998]. In a recent study, white matter damage (e.g., in the posterior limb of internal capsule) observed at three months (but not at birth) in patients with NAIS was associated with unilateral motor deficits later in life [van der Aa et al., 2013]. In chronic paediatric stroke, CST atrophy notably in the cerebral peduncle has been demonstrated to correlate with hemiparesis severity [Bouza et al., 1994; Domi et al., 2009].

The central role of the CST in motor outcome, independently of lesion size has increasingly been recognized in adult stroke [Liepert et al., 2005; Schaechter et al., 2008; Wenzelburger et al., 2005]. While simple measures of lesion volume did not seem to predict motor recovery [Jayaram et al., 2012; Puig et al., 2011], the overlap between the infarct and the CST, and the fractional anisotropy in the CST were shown to be related to upper limb motor impairment at the chronic stage of adult stroke [Lindenberg et al., 2010; Riley et al., 2011; Zhu et al., 2010]. One way to examine the microstructural change of white matter projections in vivo is diffusion‐weighted MRI, which measures the diffusion of water molecules within tissue as a surrogate of white matter orientation and integrity [Jones, 2008]. Studies showed that loss in the microstructural integrity of CST projections is associated with deficits in motor function (for a review, see Scheck et al. [Scheck et al., 2012]). This has also been demonstrated in acute phase following NAIS [van der Aa et al., 2011] but to date never described in later stage.

In our study, changes in microstructure related to CP were detected in the CST both at the level of the internal capsule and of the CSO as indicated by MD increase. Other diffusion parameters changes were mainly driven by increased RD leading to decreased FA. This has been previously explained by loss of integrity of the axolemma and/or myelin sheath (e.g., [Beaulieu, 2002; Schaechter et al., 2009]). These changes were even found when excluding the subjects with visible lesions in the internal capsule, underlining that DWI is highly sensitive to microstructural tissue damage in the CST [Groeschel et al., 2014]. The interpretation of changes in DTI parameter in the CSO, however, is more challenging as it comprises an area with complex underlying crossing fibers. One image voxel in the CSO contains both association, commissural and projection fibers [Jeurissen et al., 2013]. In this complex crossing‐fiber region, tissue damage can result in lower or higher FA [Douaud et al., 2011; Groeschel et al., 2014] and axial and radial diffusivities may not be meaningful [Wheeler‐Kingshott and Cercignani, 2009]. Still, we found significant microstructural tissue damage in this region in the CP group as detected by diffusion parameters (MD, AD, and RD).

In this study, both VSLM and diffusion analyses converged to the observation that, in addition to cortical lesions, loss in the integrity of the white matter at the level of the CSO was highly associated with the presence of unilateral CP. To date this is the first description of the implication of this in motor impairment after neonatal stroke. Going more precisely into the study of the CST using lesion categorical‐analysis in acute stroke adults, Shelton and Reding [Shelton and Reding, 2001] demonstrated that motor function worsened as the lesion location progressed from cortex to corona radiata to internal capsule. The junction of the corona radiata and the CST was also identified as the brain area that correlated most with poor hand motor ability in a recent VLSM study in 47 adults [Lo et al., 2010]. Centrum semiovale infarcts have been shown particularly detrimental to motor function in adults [Gauthier et al., 2009], and the only consistently replicated predictors of residual motor deficit was a lesion location at the intersection of the corona radiata and descending pyramidal fibers [Lindenberg et al., 2010].This region corresponds to the fanning of the main fibers originating from the premotor, motor, and sensory cortices as they converge to form the CST bundle, cranially to the internal capsule. The integrity of this convergence zone may be critical for the motor outcome, as it contains redundant tracts and potential supports for the recovering process, which may be rendered unusable by the injury, as elegantly discussed elsewhere [Shelton and Reding, 2001]. Lesions of this area could be due either to Wallerian degeneration of fibers originating from the injured cortex in case of a large cortical infarct or to direct CSO ischemia as described in adults [Beckmann et al., 2010; Bogousslavsky and Regli, 1992] and related to occlusions of the insular perforating arteries [Delion et al., 2015].

In this context, comparison with neonatal imaging can shed light on the exact mechanism of the lesions in this area. When visually inspecting our heterogeneous neonatal data, we found an obvious early involvement of the CSO in 12 of the 14 subjects who later developed unilateral CP (e.g., see Fig. 6). These results argue for an ischemic insult in this region. We did not find an isolated infarct of the CSO in our cohort in line with results of Ecury‐Goossen et al. [Ecury‐Goossen et al., 2013]. While isolated lesions in CSO can be described in patients with CP, these lesions can be confused with punctuate lesion in the white matter and should not be considered as arterial infarct [Ecury‐Goossen et al., 2013].

Taken globally our results do not advocate for highly efficient motor plasticity after NAIS. Indeed we did not observe preserved or nearly preserved motor function in children with lesions in the primary motor system as would have been the case if efficient motor function was to reorganize in other anatomical regions, including the contralateral motor cortex. All children with lesions in the primary motor system (precentral gyrus, CST, and basal ganglia) developed unilateral CP at the age of 7. Moreover, we found that the anatomical circuits related to CP were very close to those described in adult strokes, both in terms of cortical and subcortical territories. The close relationship between lesion localization and motor outcome in childhood ask by itself whether efficient neural plasticity can take place after NAIS in order to allow a full motor recovery.

This relatively weak plasticity may be related to the early maturation of the motor system already before and after birth as shown by several studies with PET imaging [Chugani and Phelps, 1986] or more recently with diffusion tensor imaging [Dubois et al., 2008], where the CST was shown to be the most mature amongst 11 reconstructed fiber tracts in three month old healthy infants, while for example, the arcuate fasciculus, sustaining the future acquisition of language, was amongst the least mature tracts at this age. The reduced plasticity of the motor system compared to language should thus be considered in line with the well‐established fact that the more mature a system is the less plastic it is.

Still, motor outcome seems more favourable in paediatric stroke than in adult stroke population [deVeber et al., 2000; Hendricks et al., 2002; Kim et al., 2009]. Motor plasticity is deemed higher in children as compared to adults as described in primates with better motor outcome after unilateral motor cortex injury at a younger age [Kennard, 1936]. However, pure anatomical correlates of neurological outcome, used here, do not enable a full understanding of infarct effects on motor system plasticity, when considered as a “functional network.” Multimodal imaging including functional MRI may help to better approach the functional plasticity of the motor system.

LIMITATIONS

Conclusions drawn from lesion mapping analysis depend on the accuracy of infarct location delineation and spatial categorization as cortical, subcortical or in the internal capsule. Also, lesion delineation was done using an expert consensus approach, inherently introducing a certain subjectivity. For the purpose of this study, however, it might be the most reliable approach. In addition, measures of lesion overlap across patients may be insufficient to identify reliable brain‐behaviour relationships when multiple lesion sites are potentially implicated in the same function [Godefroy et al., 1998]. Compared to this, the VLSM approach allows a more precise analysis and overcomes many of the limitations by using lesion characterization studies. However, VLSM analysis also has significant drawbacks. Diminished structural integrity of brain structures remote from the lesion is not taken into account in VLSM methods. Indeed VLSM only analyses areas defined as lesional in at least 5% of the participants (here 2/38). Therefore, conclusions cannot be drawn about areas that are not analyzed but might be relevant for a specific function. For example, remote areas showing Wallerian degeneration outside the lesion might contribute to unilateral CP following NAIS in a similar way to the description in chronic stroke adults [Gauthier et al., 2012]. Therefore, including the microstructural analysis of the CST in this study helped to identify brain damage beyond the lesion boundaries. A whole‐brain analysis might reveal additional involvement of brain networks. However, this was not the primary aim in this study.

Furthermore, our motor status classification can be considered simplistic (i.e., CP or not CP). Some patients may have an imbalanced involvement of superior or inferior limbs (see Table 1) depending on their lesion location. Although our cohort is the biggest prospective one to date of children with NAIS followed in childhood, we did not have enough patients to stratify subjects on more subtle neuromotor assessments.

CONCLUSION

NAIS is a severe event occurring usually after an uneventful pregnancy. After the first days of life, infants after NAIS are stable and parents question the long term motor outcome. Because of lack of studies in childhood on correlating imaging data and motor outcomes following NAIS, the treatment concepts and rehabilitation strategies are still not evidence‐based [Grunt et al., 2015]. In this context, our work emphasizes that long‐term motor outcome is closely linked to the extension of the infarct into cortical and subcortical regions classically considered as part of the motor system. Interestingly, loss of integrity of the CST in centrum semiovale and internal capsule was a highly reproducible marker of motor deficit following NAIS.