Abstract
Introduction
Cerebral palsy (CP) is a group of permanent, non-progressive, neurological disorders affecting movement, coordination and posture, appearing in early childhood. Around 2 per 1000 children are born with CP, making it one of the most common causes of physical disability in children (Oskoui et al., 2013). There are three clinical types of CP: spastic, dyskinetic and ataxic.
CP is caused by brain damage occurring during early development. Damage can occur pre-, peri- or postnatally. Examples of causes are maternal rubella or cytomegalovirus (CMV) infections, bacterial meningitis and complications related to preterm birth, such as asphyxia or head trauma. Many risk factors for CP have been identified, for instance low birth weight, premature birth and maternal thyroid problems (Odding et al., 2006).
Diagnosing CP usually requires a multidisciplinary approach. Along with clinical observations and developmental analysis, conventional MRI is used for the diagnosis of CP. These imaging techniques can reveal the underlying cause, such as periventricular leukomalacia (PVL), intracranial hemorrhage or damage to the basal ganglia. However, not all patients with PVL show motor dysfunction, preterm children may have PVL without showing signs of CP (Lee et al., 2011). Furthermore, 14-17% of children with functional impairment, show no abnormalities on conventional MRI (Scheck et al., 2012).
Recently, more advanced neuroimaging techniques have been used to investigate the structural and functional cerebral networks affected in CP. Structural connectivity consists of anatomical organization of the brain, whilst functional connectivity is defined as the temporal dependence of neuronal activity patterns of anatomically separated brain regions(Guye et al., 2008)(van den Heuvel et al., 2010).
Structural brain networks can be visualised using diffusion tensor imaging (DTI), based on diffusion weighted MRI. DTI can be used to map white matter tractography in the brain by looking at the motion of water molecules, from which the magnitude of diffusion (mean diffusivity, MD) as well as the direction (fractional anisotropy, FA) can be derived (Guye et al., 2008). Functional networks can be studied using functional MRI (fMRI), electroencephalography (EEG) or magnetoencephalography (MEG). fMRI shows neuronal activity by measuring Blood Oxygenation Level Dependent (BOLD) signals. EEG measures spontaneous electrical activation of the brain using electrodes placed on the scalp. MEG measures magnetic fields naturally generated by electrical activation the brain. Alongside these techniques, Connectome studies can be used to study structural and functional connectivity between brain areas. A Connectome is a map showing structural and functional neural connections in the brain (Cao et al., 2016).
Studies using these advanced imaging techniques have shown that structural and functional networks develop in a specific order. Structural networks have been observed to show almost adult-like organization in the postnatal period. Functional networks undergo more drastic changes during development (Cao et al., 2016). The aim of this study is to systematically review what is known about these structural and functional cerebral networks in patients with cerebral palsy.
Research question
What is known about the structural and functional brain networks in patients with cerebral palsy?
Methods
A PubMed search was conducted on 25th of July 2017. Terms used in the initial search were: ‘cerebral palsy’, ‘MRI’ (magnetic resonance imaging), ‘fMRI’ (functional magnetic resonance imaging), ‘DTI’ (diffusion tensor imaging), ‘EEG’ (electroencephalography), ‘MEG’ (magnetoencephalography) and ‘connectome’. These terms were combined using the ‘AND’/’OR’ function. This resulted in 2370 articles. To reduce the amount of articles, MeSH terms were used for cerebral palsy, MRI, fMRI and EEG. This reduced the number of articles to 893.
Next, the ‘NOT’ function was used to rule out irrelevant studies. Excluded terms were ‘epilepsy*’, ‘seizure*’, ‘rehabilitation*’, ‘therapy*’, ‘treatment*’, ‘muscle*’, ‘classification*’ and ‘prognosis*’. By excluding these terms the amount of relevant articles was brought down to 335. An example of the search strategy is provided in the appendix (appendix 1).
The relevance of articles was judged by appraising their title and abstract. Studies were selected upon meeting criteria for eligibility (appendix 2). Studies on cerebral palsy, using at least one of the following imaging techniques: fMRI, DTI, EEG, MEG and/or Connectome were considered. Reasons for exclusion were studies written in another language than English, studies of which a full-text version could not be obtained and studies conducted before the year 2000. Furthermore, duplicates were removed. Articles deemed relevant were read full-text.
A selection of 8 articles was included after reading full-text versions. A snowball search was conducted by appraising the references of the 8 selected articles. This lead to a total of 15 articles used for this review, one of them being a systematic review.
Figure 1. Study selection
Results
The 14 selected empirical studies used the following imaging techniques: fMRI, DTI or EEG. No relevant articles using MEG or Connectome were found. 12 studies researched structural networks using DTI (Bleyenheuft et al., Ceschin et al., Glenn et al., Hoon et al., Koerte et al., Lee et al., Mu et al., Muramaki et al., Nagae et al., Son et al., Thomas et al and Yoshida et al.). 3 studies researched functional networks; 2 studies using fMRI (Burton et al. and Lee at al.), 1 using EEG (Kulak et al.). In 12 out of 14 studies, conventional MRI findings of patients were known. Study characteristics are summarized in the appendix (appendix 3).
Structural networks
The 12 studies using DTI included a total of 220 patients. The majority of patients had spastic diplegic CP (SDCP) (66.8%). Other types of CP in the patient group were hemiplegia, paraplegia, quadriplegia and athetotic CP. Conventional MRI findings were known of all patients. Patients can be divided into 3 groups by their MRI results: PVL group (70.5%), no abnormalities group (11.4%) and other abnormalities than PVL group (18.2%).
Figure 2. Conventional MRI results of patients assessed with DTI
White matter tracts of 43 SDCP patients in the PVL group were mapped (Lee et al., 2011). These patients had PVL, no other structural abnormalities were seen on conventional MRI. Reduced FA was seen within almost all white matter tracts (i.e. internal capsule, external capsule, cerebellar peduncle, corpus callosum), posterior tracts being more affected than anterior tracts.
In another 28 PVL patients (75% SDCP) (Hoon et al., 2009), mapping of the white matter tracts revealed 27 patients (96,4%) had posterior thalamic radiation (PTR) injury. 8 patients (28,6%) showed injury to the corticospinal tract (CST).
17 SDCP patients in the PVL group (Ceschin et al., 2015) showed reduced FA in the optic radiation, inferior fronto-occipital fasciculi (IFOF), inferior longitudinal fasciculi (ILF), limbic system (cingulum, fimbria, fornix), frontal lobe and motor system (corona radiata, corpus callosum, capsula interna, cerebellar peduncles) on DTI. Abnormal mean diffusivity was found in the optic radiation, posterior thalamic radiation, superior longitudinal fasciculi (SLF), ILF and IFOF. In the optic radiations and posterior thalamic radiation MD was decreased.
24 CP patients in this group were consecutively scanned in an ongoing study (75% spastic diplegia) (Nagae et al., 2007). 19 white matter tracts were scored as abnormal on color coded DTI maps. Among affected regions were the CST, posterior limb and retrolenticular part of the internal capsule, PTR, SLF, ILF, IFOF, corona radiata, cerebellar peduncles and corpus callosum.
DTI fibre count was performed in 5 SDCP patients (Thomas et al.). In addition, FA and MD values were calculated. The fibre count was reduced in the CST, corticobulbar tract (CBT) and superior thalamic radiation. This corresponded with increased MD in these areas, along with the thalamus. FA was reduced in the CST and corticobulbar tract (CBT). FA was also reduced in the thalamus, basal ganglia (caudate nucleus and lentiform nucleus) and corpus callosum.
FA of the transcallosal motor fibres (TCMF) and CST was calculated of 7 SDCP patients in the PVL group (Koerte et al., 2011). No other pathologies than PVL were found on conventional MRI. The CP patients showed lower FA values of the TCMF in comparison to the healthy control group. No difference in FA was found regarding the CST between the 2 groups.
10 patients with PVL, of which 5 with CP (spastic paraplegia and quadriplegia) and 5 with normal motor development were compared using DTI (Murakami et al., 2008). Significant differences in FA were found in motor tracts, the values in the CP group were lower than those in children with normal motor function.
45 CP patients (19 athetotic and 26 spastic diplegic) were assessed. FA and MD were calculated for 22 brain regions (Yoshida et al., 2011). In the athetotic CP group loss of white matter volume (26%), rolandic type injury (26%), PVL (21%), parasagittal cerebral injury (11%) and thinning of the corpus callosum (5%) were found on conventional MRI. 2 patients (11%) displayed no abnormalities on conventional MRI. In the spastic CP group, 88% of patients showed PVL on conventional MRI. 2 patients had loss of white matter volume, 1 patient showed abnormal T2 hyperintensity in the putamen. Overall, the CP groups had reduced FA in comparison to the control group. The athetotic CP group showed reduction in 91% of the investigated brain regions; bilateral putamen, globus pallidus, bilateral thalamus, bilateral posterior limb of internal capsule, bilateral posterior thalamic radiation, bilateral anterior limb of internal capsule, bilateral SLF, ILF, bilateral cingulum, bilateral CST and corpus callosum. In the spastic CP group, FA was reduced in 32% of the investigated regions; bilateral posterior limb of internal capsule, bilateral posterior thalamic radiation, bilateral SLF and splenium of corpus callosum. MD had an overall tendency to be increased. Patients with athetotic CP showed increased MD in 18 locations; bilateral globus pallidus, bilateral thalamus, bilateral posterior limb of the internal capsule, bilateral anterior limb of the internal capsule, posterior thalamic radiation, SLF, cingulum, genu and splenium of corpus callosum. In the spastic CP group increased MD was found in 3 white matter regions; posterior thalamic radiation, SLF and splenium. Abnormalities in the spastic CP group were confined to white matter, whereas the athetotic CP group also showed abnormalities in gray matter.
Corticospinal tract areas were measured in 12 patients with hemiplegia (Bleyenheuft et al., 2007). On conventional MRI, asymmetry of the CST and cerebellar peduncles was seen. The contralateral structures were significantly smaller than those on the ipsilateral side. Control group showed no asymmetry between the CST and peduncles. Using DTI, the same asymmetry was observed in all subjects with hemiplegia. Symmetry indexes of the CST and cerebellar peduncles were calculated ((contralateral area/ipsilateral area) x 100). Smaller symmetry indexes were found by using DTI, suggesting conventional MRI might underestimate the asymmetry.
Diffusion parameters were measured in 15 patients with hemiplegia (Glenn et al., 2007). On conventional MRI, various types of lesions were seen, among them polymicrogyria and perinatal infarction. 3 patients showed no abnormalities on conventional MRI. None of the patients had PVL. FA and MD were calculated for the contra- and ipsilateral pyramidal tracts. Patients with more severe clinical symptoms of hemiplegia had increased FA asymmetry. FA values in the contralateral side were decreased. MD was increased in the affected pyramidal tract in the patient group compared to the control group. Again, the severity of hemiplegia correlated with the MD values. Patients with mild hemiplegia did not show significantly increased MD in comparison to the control group.
16 occult spastic diplegic CP patients showed no abnormalities on conventional MRI (Mu et al., 2014) . DTI revealed these patients to have significantly decreased fractional anisotropy (FA) compared to the healthy control group. This appeared in several white matter regions; bilateral white matter tracts in the prefrontal lobe, temporal lobe, internal and external capsule, thalamus, cingulum, corpus callosum, cerebellum and brainstem.
CST of 4 patients with hemiparetic CP was studied using DTI (Son et al. 2007). All patients showed no focal lesions on conventional MRI. FA and MD were measured, and asymmetry indexes were estimated. In the patient group decreased FA values were found in the corona radiata, internal capsule, cerebral peduncle Additionally, higher values for the MD asymmetry index were found in the CP group.
Functional networks
3 studies investigated functional connectivity, 2 using fMRI (Burton et al. and Lee et al.), 1 using EEG (Kulak et al.), including a total of 37 CP patients. 25 patients had spastic diplegia, 12 patients had hemiplegia. Conventional MRI results were known of 13 patients.
Resting state fMRI was performed on 13 subjects with spastic diplegic CP, data was obtained from 11 (Lee et al., 2011). All subjects showed PVL on conventional MRI. Patients were instructed to rest without movement and keep their eyes closed for 5 minutes during the scan, attempting to sleep. Motor cortex and thalamic connectivity were analysed. The motor cortex in the healthy control group showed broad functional connection to the somatosensory cortex, visual cortex and cerebellum. The patient group displayed decreased connectivity within the visual cortex, superior and inferior parietal lobules, cingulate motor area, presupplementary motor area, paracentral lobule and somatosensory cortex (p<0.005). However, the motor cortex connections to the adjacent parietal area were more expanded in the patient group. In healthy controls, the thalamus showed connectivity to the basal ganglia, cerebellum and cingulate cortex. In patients with CP, the connectivity from the thalamus to the caudate nucleus, cingulate cortex and cerebellum was reduced (p<0.005). Along with these findings, severe optic radiation damage was seen.
fMRI was performed on 12 subjects with spastic diplegic CP (Burton et al., 2010). Conventional MRI findings in these patients were not mentioned. Patients were blindfolded, and instructed to keep their eyes open and stay awake. No task was performed during the scan. Functional networks connected to the parietal somatosensory areas were studied. Overall, CP patients showed expanded networks in comparison to healthy control groups. For example, in the control group for the left Brodmann area 1 (LBA1) region of interest (ROI), the networks linked to the parietal somatosensory areas consisted of middle of the pre- and postcentral gyri, small part of parietal operculum and the frontal premotor cortex. In patients with spastic diplegia the network enclosed all of the pre- and postcentral gyri, fragment of the superior temporal plain, medial frontal cingulate cortex and the inferior supramarginal cortex. For the left OP1 ROI in healthy controls the linked networks consisted of the parietal and temporal operculum, insula, inferior supramarginal gyrus and some parts of the left postcentral gyrus. In diplegia patients, the network existed of the post- and precentral gyri, temporal opercula, parietal opercula, insula and part of the adjacent cingulate cortex. Further statistical analysis also backed that diplegia patients had larger networks. Other regions where connectivity was significantly larger in CP patients than controls include LBA3, RBA3, LBA2, RBA2, LBA5, LBA7 and RBA7.
An EEG was made of 12 patients with right hemiparetic cerebral palsy (Kulak et al., 2005). A CT scan was performed; 7 patients showed ventricular asymmetry, 2 showed porencephaly, 2 showed signs of an infarct and 1 patients had PVL. The EEG was performed in resting state using 14 scalp electrodes, patients were instructed to lie down with their eyes closed. No task was performed during the EEG. Interhemispheric coherence (ICoh) and intrahemispheric coherence (HCoh) were calculated, these values are interpreted as reflecting levels of functional connectivity between brain regions. In the alpha band, HCP patients appeared to have significantly lower ICoh values compared to the control group between several electrode pairs; C3-C4, O1-O2, P3-P4, T3-T5 and T4-T6. T3-T5 and T4-T6 having the lowest detected values. In the theta band, values were significantly lower in C3-C4, P3-P4 and O1-O2 and higher in F3-F4, F7-F8, T3-T5 and T4-T6. In the delta band, lower values compared to the control group were detected in F3-F4, C3-C4 and O1-O2. ……