Article Text

Download PDFPDF

DTI reveals network injury in perinatal stroke
  1. Jeroen Dudink1,2,
  2. Serena J Counsell3,
  3. Maarten H Lequin2,
  4. Paul P Govaert1,4
  1. 1Department of Neonatology, Sophia Children's Hospital, Rotterdam, The Netherlands
  2. 2Pediatric Radiology, ErasmusMC Sophia, Rotterdam, The Netherlands
  3. 3Imaging Sciences Department, MRC Hammersmith/St Mary's Comprehensive Biomedical Research Centre, London, UK
  4. 4Department of Pediatrics, Section of Neonatology, Paola Kinderziekenhuis, Antwerp, Belgium
  1. Correspondence to Dr Jeroen Dudink, Department of Neonatology, Sophia Children's Hospital, Dr Molenwaterplein 60, Rotterdam 3015GJ, The Netherlands; j.dudink{at}


Background Previous research showed acute diffusion-weighted imaging changes in pulvinar after extensive cortical injury from neonatal stroke. The authors used diffusion tensor imaging (DTI) to see how separate regions of ipsilateral thalamus are directly affected after a primary hit to their connected cortex in neonatal stroke.

Methods The authors analysed DTI images of three term infants with acute unilateral cortical arterial ischaemic stroke. Probabilistic tractography was used to define separate thalamic regions of interests (ROIs). The authors evaluated the three eigenvalues (EV) and apparent diffusion coefficient (ADC) values in the ROIs.

Results The ADC and EV in voxels of ROIs placed within the nuclei corresponding to ischaemic cortex were significantly lower than those in the unaffected contralesional thalamic nuclei.

Conclusions Our findings support the concept of acute network injury in neonatal stroke. ADC and EV were altered in specific thalamic regions that corresponded to the specific cortical areas affected by the primary ischaemic injury.

View Full Text

Statistics from


MRI is the investigation of choice in perinatal stroke. Diffusion-weighted imaging (DWI) can detect neonatal ischaemic brain injury within 24 h of its onset and monitor acute changes for several days. DWI also allows to visualise secondary effects of neuronal injury on connected projection tracts. An example is the early stage of Wallerian degeneration.

Acute changes in pulvinar were found after extensive cortical injury from neonatal stroke, which could not be explained by disturbances in vascular perfusion.1

It is still unclear whether other regions of the thalamus, next to the pulvinar, also show acute changes directly after a primary hit to the connected cortex. Such a study would further support the concept of ‘acute secondary network injury’ and may lead to better understanding of the variability in neurological outcome in perinatal stroke.

Diffusion tensor imaging (DTI) is an MRI sequence with which the effects of network injury on the thalamus can be studied. It allows objective and reproducible in vivo assessment of tissue characteristics. Using a relatively new DTI postprocessing technique, ‘probabilistic tractography’ and thalamic nuclei can be mapped and DTI measurements can be made in those nuclei.2

In the study reported here, we applied this technique to evaluate the acute effects of neonatal arterial ischemic stroke (NAIS) on the separate regions of the thalamus.

Subjects and methods


DTI images of three term infants with extensive unilateral cortical NAIS were retrospectively drawn from the Erasmus MC-Sophia Children's Hospital database. These infants were born after an uncomplicated pregnancy with normal birth weight and APGAR scores of >5 after 5 min. Their presenting signs were apnoea and seizures on day 1 after birth. Patient 1 had a left posterior truncal middle cerebral artery (MCA) stroke (sensory, parietal and temporal cortex), patient 2 had a left posterior cerebral artery (PCA) stroke (occipital cortex) and patient 3 had a left complete MCA stroke (frontal, motor, sensory, parietal and temporal cortex). Long-standing antepartum injury was excluded from normal findings on routine first day ultrasound scanning. All patients underwent MRI scanning in the latter part of the first week after birth.

MRI technique

Image acquisition

Imaging was performed on a 1.5-T GE EchoSpeed scanner (GE Medical Systems, Milwaukee, WI, USA). The protocol included T1- and T2-weighted and DTI sequences (25 gradient directions, b=1000 s/mm2).

Data analysis

Data analysis was performed using FSL software tools (FSL version 4.1.4). We used the FSL automatic registration programme (FLIRT) to register the DTI data from each patient to a template image (ICBM-MNI 152) and cortical masks (Harvard Oxford Cortical Structural Atlas). We divided the cortex of both the hemispheres into functional corticothalamic areas: the frontal cortex (with exclusion of the primary motor area), the primary motor, primary sensory, parietal (primary sensory excluded), temporal and occipital. We manually delineated the thalamus on T2-weighted images and propagated them onto the infant's native DTI data using FLIRT.2

With both the delineated thalamus and the different cortical masks now registered on DTI, we used FSL probabilistic tractography software (connectivity-based seed classification) to automatically find the separate thalamic regions (nuclei). This technique of thalamocortical mapping is described by Counsell et al.2 Finally, the separate thalamic regions of both hemispheres were used to place regions of interests (ROIs) from which DTI measurements were taken: the three eigenvalues (EV) (λ1, λ2 and λ3) were measured and the apparent diffusion coefficient (ADC) was calculated (figure 1).

Figure 1

Images of a patient with a posterior truncal MCA stroke: (i) ADC map showing low (dark) signal intensity (SI) in area of stroke; (ii) same ADC-map after automated image registration software was used to co-register DTI data from the patient to standard cortical masks (Harvard Oxford Cortical Structural Atlas) (Orange= frontal cortex (minus the primary motor area), red = primary motor cortex, dark-blue = primary sensory cortex, yellow = parietal cortex (minus the primary sensory cortex) and lightblue = occipital cortex); (iii) thalamo-cortical connections were assessed for every voxel in the thalamic masks (green) using connectivity-based seed classification with predefined thresholds; yellow-red area is the seed mask of the parietal cortex, there is a clear difference between right and left thalamus (R>L); (iv) The thalamic connectivity seeds were used to place ROIs from which the DTI measurements were taken (blue squares).

Statistical method

The EV and ADC values of the voxels in the ROIs of ipsilateral thalamus were compared with those of contralateral thalamus, using the paired t-test with two tails; a p-value of <0.05 was considered statistically significant (Graphpad Instat for Macintosh).


ADC values

The paired t-test with two tails revealed significantly lower ADC values in the ROIs placed within the regions of the thalamus corresponding to the ischaemic cortex (table 1) compared with the same ROIs placed in the contralesional thalamus.

Table 1

Mean and SD ADC values (in 103×mm2/s) measured in the ROIs placed in the areas of the thalamus showing the highest connectivity with the ‘frontal’, ‘occipital’, ‘parietal’, ‘sensory’, ‘motor’ and ‘temporal’ cortex


All EV (λ1, λ2 and λ3) of the voxels in the ROIs of the affected ipsilateral thalamic regions were significantly lower than those in the corresponding contralesional regions. Comparing mean differences of λ1, λ2 and λ3 yielded an ‘absolute’ mean difference λ1 > λ2 > λ3 in all affected thalamic regions. Of all three, the relative mean differences (ie, the mean difference divided by the arithmetic mean) in the λ1 values were the largest.


Our patients show significant changes in DTI values of voxels in the regions of ipsilateral thalamus corresponding to the injured cortex. These changes are consistent with acute secondary network injury (specifically in thalamic regions) directly after a primary hit to the connected cortex.

For decades, pathologists have adhered to the concept that primary cortical injury can lead to cell loss in the thalamus. Global thalamic shrinkage has been associated with preterm white matter injury and DTI probabilistic tractography showed long-term influence of perinatal venous white matter infarction on thalamus in 2-year-old infants.2

Govaert et al1 recently reported findings of acute secondary effects of cell injury in regions connected to the primary lesion sites in NAIS. The subjects were term infants with cortical stroke who showed signal hyperintensities in the ipsilateral pulvinar on DWI and a decrease in ADC values during the acute phases of the injury.

Acute diffusion decrease in grey matter remote from the ischaemic area has also been reported in experimental animal models.3 The decrease seemed to be temporally related to swelling of neurons and perivascular astrocyte feet. Astrocytic and microglial activation may also account for the lower mobility of water molecules in the ipsilateral thalamus.

Several mechanisms have been suggested to explain secondary neuronal cell death in the ipsilateral thalamus.4 One of these is excitotoxic injury: cortical neurons suffering from electrographic seizures transmit glutamatergic signals to connected thalamic nuclei. Another mechanism might be the neurodegeneration in the connected thalamic regions following primary injury due to loss of trophic support from their cortical targets. This type of transsynaptic thalamic cell death is mainly apoptotic and develops over hours to days.4

In our patients, in the affected regions of the ipsilateral thalamus, the reduction in the primary (λ1) was larger than that in the second and third EV (λ2 and λ3). Correspondingly, Sorensen et al5 showed that in abnormal grey matter regions of patients with acute ischaemia, the reduction in the primary EV was the largest. Combining histopathology with DTI will shed more light on the underlying mechanisms of these findings.

Our study has limitations. One is the lack of a large cohort of normal DTI data in different thalamic subnuclei. We used the contralesional thalamus as a reference. Other groups have reported normal DTI values in thalamus, which seem to be of the same order of magnitude as the values found in our study. Nevertheless, to our knowledge, none of these studies have used probabilistic methods to measure DTI values in different thalamic subnuclei. Patient 2 had a PCA stroke and abnormal DTI findings in pulvinar. One could theoretically argue that perforator stroke in pulvinar due to occlusion of a thalamogeniculate branch caused diffusion abnormalities. However, in patients with MCA strokes, who exhibit no primary lesion in pulvinar because that nucleus is perfused from the PCA, we still showed acute changes in pulvinar similar to those in PCA strokes. Another limitation is the absence of histopathological studies of secondary thalamic lesions in term newborns.


Our DTI findings support the concept of acute network injury in NAIS, showing how deep grey matter is directly affected after a primary hit to its connected cortex. This intriguing concept may lead to better understanding of the variability in neurological outcome in perinatal stroke.

Future research efforts should be directed at determining the relevance of delineating network injury, its influence on prognosis and the impact of early treatment of a newborn with stroke.


View Abstract


  • Ethics approval The study was approved by ErasmusMC METC.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Fantoms
    Martin Ward Platt