Article Text

Download PDFPDF

Case Series
Neuroimaging selection for thrombectomy in pediatric stroke: a single-center experience
  1. Sarah Lee1,2,
  2. Jeremy J Heit3,
  3. Gregory W Albers1,
  4. Max Wintermark3,
  5. Bin Jiang3,
  6. Eric Bernier1,
  7. Nancy J Fischbein3,
  8. Michael Mlynash1,
  9. Michael P Marks3,
  10. Huy M Do3,4,
  11. Robert L Dodd3,4
  1. 1 Stanford Stroke Center, Department of Neurology & Neurological Sciences, Stanford University School of Medicine, Stanford, CA, USA
  2. 2 Division of Child Neurology, Department of Neurology & Neurological Sciences, Stanford University School of Medicine, Stanford, CA, USA
  3. 3 Department of Radiology, Division of Neuroimaging & Neurointervention, Stanford University School of Medicine, Stanford, CA, USA
  4. 4 Department of Neurosurgery, Stanford University School of Medicine, Stanford, CA, USA
  1. Correspondence to Dr. Sarah Lee; slee10{at}


Background The extended time window for endovascular therapy in adult stroke represents an opportunity for stroke treatment in children for whom diagnosis may be delayed. However, selection criteria for pediatric thrombectomy has not been defined.

Methods We performed a retrospective cohort study of patients aged <18 years presenting within 24 hours of acute large vessel occlusion. Patient consent was waived by our institutional IRB. Patient data derived from our institutional stroke database was compared between patients with good and poor outcome using Fisher’s exact test, t-test, or Mann-Whitney U-test.

Results Twelve children were included: 8/12 (66.7%) were female, mean age 9.7±5.0 years, median National Institutes of Health Stroke Scale (NIHSS) 11.5 (IQR 10–14). Stroke etiology was cardioembolic in 75%, dissection in 16.7%, and cryptogenic in 8.3%. For 2/5 with perfusion imaging, Tmax >4 s appeared to better correlate with NIHSS. Nine patients (75%) were treated: seven underwent thrombectomy alone; one received IV alteplase and thrombectomy, and one received IV alteplase alone. Favorable outcome was achieved in 78% of treated patients versus 0% of untreated patients (P=0.018). All untreated patients had poor outcome, with death (n=2) or severe disability (n=1) at follow-up. Among treated patients, older children (12.8±2.9 vs 4.2±5.0 years, P=0.014) and children presenting as outpatient (100% vs 0%, P=0.028) appeared to have better outcomes.

Conclusions Perfusion imaging is feasible in pediatric stroke and may help identify salvageable tissue in extended time windows, though penumbral thresholds may differ from adult values. Further studies are needed to define criteria for thrombectomy in this unique population.

  • pediatrics
  • thrombectomy
  • embolic
  • MRI
  • Mr perfusion

Statistics from

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.


Arterial ischemic stroke in children is not rare, and may result in devastating, lifelong disability.1 Unfortunately, diagnosis is often delayed due to lack of awareness and challenges obtaining urgent imaging in children.2 The extended 24 hours time window for thrombectomy in well-selected adult patients3 4 represents a major opportunity for acute stroke treatment in children, for whom diagnosis within 4.5 or 6 hours is often unrealistic.5 However, no randomized trials currently exist to determine how best to select pediatric patients for thrombectomy, or whether it is beneficial. We report 12 pediatric patients presenting to our institution with large vessel occlusion, discuss demographic, clinical and neuroimaging considerations that may be used to help determine thrombectomy eligibility in children, and report outcomes for both treated and untreated patients.


We performed a retrospective cohort study of all patients <18 years of age presenting to our institution with acute stroke symptoms and large vessel occlusion who were considered for thrombectomy between 2008 and 2018. Patient consent was waived by our institutional IRB.

Clinical, demographic, and imaging data were derived from our institutional prospectively-maintained stroke and thrombectomy databases. As part of our established pediatric stroke code process, a National Institutes of Health Stroke Scale (NIHSS) or pediatric NIHSS6 was documented at baseline.

Magnetic resonance imaging (MRI) was our preferred imaging modality, and our ‘pediatric quick stroke protocol’ includes the following sequences: localization (0:14 min), diffusion-weighted imaging (DWI; 0:40 min), gradient-recalled echo (GRE; 1:44 min), T2 fluid-attenuated inversion recovery (FLAIR; 2:25 min); time-of-flight magnetic resonance angiography (MRA; 4:23 min), arterial spin labeling (ASL; 4:28 min), and MR perfusion (perfusion-weighted imaging (PWI); 5:23 min). MR perfusion was performed using a dynamic-contrast susceptibility imaging (DSC) technique following injection of a gadolinium contrast agent (MultiHance, Bracco, Milan, Italy) into an antecubital vein. MR perfusion images were processed using automated software (RAPID, iSchemaView, Menlo Park, CA, USA).

Patients who could not undergo MRI underwent baseline imaging by computed tomography (CT) with CT angiography (CTA) of the brain, with or without CT perfusion. CT perfusion was also processed using RAPID software. Non-contrast CTs were independently scored for ischemia using the Alberta Stroke Programme Early CT Score (ASPECTS) by two neuroradiologists (MW/JJH). Inter-rater discrepancies were adjudicated by a third neuroradiologist (NJF).

Thrombectomy was generally considered in patients with the following profile: (1) an acceptably small core infarct volume, typically <70 mL quantified by DWI, or ASPECTS ≥7 on noncontrast CT; (2) evidence of a large vessel occlusion (internal carotid artery, first or second part of the middle cerebral artery, vertebral artery, or basilar artery) on MRA or CTA; (3) when available, evidence of salvageable penumbra determined by MR or CT perfusion deficit on the time-to-maximum delay of tissue residue (Tmax) maps; (4) when available, disabling neurologic exam with cortical signs, typically NIHSS ≥6.

Thrombectomy device type, technique, and adverse events were noted; all devices used were US Food and Drug Administration (FDA)-cleared for adult thrombectomy. Reperfusion was graded with the modified Thrombolysis in Cerebral Infarction (TICI) score7 by the treating neurointerventional radiologist.

Radiographic outcome measures included vessel recanalization, final stroke volume, and presence of hemorrhagic transformation. Clinical outcomes were assessed with the Pediatric Stroke Outcome Measure (PSOM), with good outcome defined as PSOM severity score 0–1 (no or mild deficits), and poor outcome defined as PSOM severity score 2–3 (moderate-severe deficits) or death8 (see online supplementary material). If PSOM was not documented at follow-up clinic visits, it was scored retrospectively by a pediatric stroke neurologist (SL). Baseline demographic, clinical, and radiographic data were compared between patients with good and poor outcome using Fisher’s exact test, t-test or Mann-Whitney U-test.


Twelve children presented to our institution between 2008 and 2018 within 24 hours of last known well with stroke symptoms attributed to a large vessel occlusion (table 1). For detailed case information, please see online supplementary material.

Table 1

Clinical and imaging characteristics.

Eight patients (66.7%) were female, mean age was 9.7±5.0 years (range 8 months–16 years). Nine cases had a documented NIHSS at presentation; median NIHSS was 11.5 (IQR 10–14). Known pre-existing stroke risk factors were present in 9/12 patients (75%): eight had congenital heart disease (two on mechanical circulatory support), and one had both patent foramen ovale and migraine with aura.

Five patients (41.7%) were evaluated by CT/CTA; one teenager also had CT perfusion. Median ASPECTS score was 8 (IQR 6.5–8.5). Seven patients (58.3%) underwent MRI/MRA, five with RAPID perfusion (5/5 technically adequate) and six with ASL (4/6 technically adequate). Vessel occlusion involved the anterior circulation in 10/12 patients (83.3%). For patients with anterior circulation stroke who underwent MRI, median initial core infarct volume was 30.9 mL (IQR 19.2–35.2). For those who also had DSC perfusion, median penumbra volume was 23.5 mL (IQR 12.0–36.5) for Tmax >6 seconds, and 60.0 mL (IQR 55.0–93.0) for Tmax >4 seconds.

Of the five patients who had RAPID perfusion imaging, a Tmax threshold >4 seemed to better represent critical hypoperfusion. For patient #3, RAPID did not detect Tmax >6, but a significant volume of Tmax >4, which corresponded with global aphasia on exam (figure 1). For patient #5, minimal subcortical tissue was estimated at Tmax >6, but a significant volume at Tmax >4, which corresponded with the patient’s forced gaze deviation and profound neglect (figure 2). Interestingly, for patient #2, a significant amount of Tmax >6 was detected, but of note this child presented much later (treated at 19 hours after last known well, versus ~6 hours for patients #3 and #5). By contrast, patient #10 demonstrated a largely matched lesion, and thrombectomy was not pursued (figure 3a-b). Of the four patients with both technically adequate ASL and RAPID perfusion, the ASL lesion did appear to correlate with the Tmax >4 volume.

Figure 1

Patient #3: Initial magnetic resonance imaging (MRI) performed ~4.5 hours after last known well shows subcortical infarction and proximal M2 cut-off. No tissue estimated at Tmax >6 seconds, but 38 mL of Tmax >4 seconds corresponding with arterial spin labeling hypoperfusion and aphasia on exam. The patient underwent thrombectomy with Thrombolysis in Cerebral Infarction (TICI) 2b reperfusion. ADC, apparent diffusion coefficient.

Figure 2

Patient #5: Ischemic core of 16 mL, minimal tissue estimated at Tmax >6 seconds, but significant volume at Tmax >4 seconds corresponding with arterial spin labeling hypoperfusion and neglect and gaze preference on exam. The patient was taken for thrombectomy, with Thrombolysis in Cerebral Infarction (TICI) 3 reperfusion. ADC, apparent diffusion coefficient.

Figure 3

(A) Patient #2 was found down with aphasia and right-sided weakness. Magnetic resonance angiography (MRA) showed complete occlusion of left ICA terminus; RAPID map demonstrated a large area of tissue at risk. The patient was taken for thrombectomy with Thrombolysis in Cerebral Infarction (TICI) 3 reperfusion. (B) Patient #10 with iatrogenic carotid dissection during a medical procedure initially had no associated neurologic deficits, then developed acute onset left hemiparesis. MRA showed new tandem M1 occlusion; RAPID map with largely matched lesion with small volume of tissue at risk primarily over a non-eloquent area. Thrombectomy was not pursued.  ADC, apparent diffusion coefficient; DWI, diffusion-weighted imaging .

Of the 12 patients evaluated, nine (75%) were treated: seven patients underwent thrombectomy alone, one received both IV tissue plasminogen activator (tPA) and thrombectomy, and one received IV tPA and was taken for thrombectomy, but initial diagnostic angiogram demonstrated recanalization, so retrieval was not performed. Of the 10 patients who did not receive IV tPA, six presented within 4.5 hours of symptom onset but had contraindications to intravenous thrombolysis, including recent hemorrhage (n=3), postoperative status (n=1), anticoagulation (n=2), and malignant-sized infarction (n=1).

Of the eight patients who underwent thrombectomy, seven (87.5%) had successful reperfusion (TICI 2b, 2c, or 3). Median final core infarct volume (for anterior circulation strokes, based on follow-up MRI) was 37.6 mL (IQR 16.0–44.3), and median core infarct growth was 10.0 (IQR 1.0–23.4). Clinically significant intracranial hemorrhage occurred in one patient who required full-dose anticoagulation for persistent clotting on extracorporeal membrane oxygenation (ECMO).

Favorable clinical outcome was achieved in 78% of treated patients (including the patient with TICI 1 reperfusion) versus 0% of untreated patients (P=0.018), and 75% of successfully reperfused patients versus 25% of non-reperfused patients (P=0.222). All three patients who did not undergo thrombolysis or thrombectomy had poor outcome, with either severe disability (one patient) or death (two patients) at 90 days. Among treated patients, older children (12.8±2.9 vs 4.2±5.0, P=0.014) and children presenting as an outpatient (100% vs 0%, P=0.028) seemed to have better outcome. No difference was detected in initial NIHSS, core infarct volume, final infarct volume, infarct growth, or stroke etiology between patients with good and poor outcome, though the sample size limits drawing definitive conclusions.


Criteria for eligibility in recent late-window adult thrombectomy trials included the presence of an anterior circulation large vessel occlusion, disabling stroke symptoms, and relatively small core infarct volume (estimated by MRI-DWI volume or cerebral blood flow <30% on CT perfusion);4 DEFUSE 3 also required neuroimaging evidence of a core-penumbra mismatch, with critically hypoperfused tissue defined by regional CT or MR perfusion Tmax >6 seconds.3 Mismatch is not often assessed in children with large vessel occlusion, either because MRI is contraindicated or unavailable and CT perfusion is discouraged, or because MR perfusion is not performed. Furthermore, perfusion thresholds for children are not established; our data suggest they may differ from adult thresholds. An increasing number of case series9–12 suggest that thrombectomy is feasible, safe, and may lead to better outcomes in children with large vessel occlusion; however, selection criteria are seldom discussed and publication bias confounds generalizability of these results. The Royal College of Paediatric and Child Health,13 the Australian Childhood Stroke Advisory Committee,14 and the American Heart Association15 have all issued statements or guidelines acknowledging that thrombectomy may be beneficial for selected children with large vessel occlusion while recognizing the absence of rigorous studies in this population.

The following sections highlight key criteria for endovascular selection in adults, discuss if and how they may be applied to children, and address gaps which currently exist for decision-making in pediatric stroke.


All randomized controlled trials for stroke have excluded patients <18 years of age. The only prospective trial in pediatric stroke, which aimed to determine if IV tPA was safe and feasible in children, excluded patients <2 years old, and was closed early due to lack of enrollment.16 Other pediatric institutions have instituted the same or older cutoffs for endovascular therapy in children,17 with the rationale that thrombectomy may be more technically challenging, and thus riskier, in younger children and infants due to smaller intracranial vessels, which are reported to approximate adult-sized vessels around age 5 years.18 Additionally, adult neurointerventionalists may not feel comfortable treating children due to lack of experience or large trials to support its use. Our series includes a 7 kg infant: technical challenges were encountered given this patient’s size, as vessel caliber could only accommodate the catheter without its sheath. Different endovascular techniques and safety considerations may be warranted for infants and smaller children; this requires further study.

Foregoing intervention in pediatric patients may also be justified by the notion that children recover better from stroke than adults due to brain plasticity. In fact, 65% of pediatric stroke survivors have lasting motor deficits, and up to a third will develop epilepsy.1

Neurologic exam/NIHSS

The NIHSS is a validated 15-item clinical assessment tool used to predict stroke severity and serves as a key inclusion criterion in adult trials. DAWN required an NIHSS ≥104 and DEFUSE 3 an NIHSS ≥63 for inclusion. Whether children stand to benefit from thrombectomy despite a low NIHSS is not known. Animal and human studies suggest that robust pial collaterals are present at birth and undergo rarefaction during the ageing process, and when subjected to other vascular risk factors and hypoxic states over time.19 For children in particular, healthy collaterals may initially supply penumbral tissue and transiently improve NIHSS; however, if and when each individuals’ collaterals will ultimately fail is a topic much debated in both pediatric and adult populations.

Relying on the clinical exam to risk-stratify pediatric patients is challenging because young children may be frightened or uncooperative, making it difficult to perform a formal NIHSS. Language and coordination testing in children is non-uniform, as ability will depend on developmental stage. In response to these challenges, the pediatric NIHSS was created, which has demonstrated excellent inter-rater reliability and has a modified version for children <2 years old.6 However, the pNIHSS may not be readily taught or available in pediatric hospitals or emergency rooms.  Finally, establishing a firm last seen normal time in an infant or young child presents a unique challenge, as subtle symptoms preceding the parent or caregiver’s observation of deficits cannot be entirely ruled out. In our series, most but not all patients were examinable with a documented NIHSS and confirmed time of onset; however, three patients were on some form of sedation at the time of their stroke. Education and dissemination of the pediatric NIHSS may help provide more accurate clinical assessments and help guide decision-making.

Time window

In response to the DAWN and DEFUSE 3 trials, adult stroke guidelines have extended the time window for thrombectomy to 24 hours from last known well. In one of the largest meta-analyses to date of 44 pediatric thrombectomy cases,11 time to treatment varied widely (range 1.5–72 hours); 16% were treated beyond 24 hours. All patients in our series were treated within 24 hours of new neurologic deficit or vessel occlusion. Though few conclusions can be drawn with the small number of cases, longer time to treatment did not seem to correlate with worse outcome in our series; a similar finding was observed among adult patients in DEFUSE 3. Collateral vessels have been postulated to play a central role in delaying expansion of infarcted tissue and may be a key factor in determining why one patient completes a stroke at 1 hour and another at 20 hours. It stands to reason that healthy children without prior vasculopathy or chronic hypoxia should have robust collaterals, which may supply the penumbra for some time, or even sustain the tissue indefinitely. That said, a large subset of children with stroke do have significant comorbidities; in our series, as well as others, congenital heart disease accounts for the majority of acute large vessel occlusions. Children with cyanotic heart disease have chronic cerebral hypoperfusion;20 understanding how this affects collateral development may also inform endovascular candidacy, and further investigation of the ‘collateralome’ in these at-risk patients is warranted.


Consistent with prior reports, stroke etiology in the majority of our cases was cardioembolic. All three patients who died had underlying congenital heart disease and were inpatient at the time of their stroke; two were on mechanical circulatory support (MCS). For children with congenital heart disease, MCS is a lifesaving option for some who previously may not have survived past infancy; however, it also confers a very high risk of stroke, both ischemic (secondary to pump thrombosis) and hemorrhagic (given the need for aggressive antithrombotic therapy).21 Thrombectomy has been successfully performed in adult patients on MCS,22 but few cases have been reported in children.23 24 Sun et al 25 reported a case of thrombectomy for an infant on ECMO who had good radiographic outcome but poor clinical outcome. Because these patients have an especially high incidence of stroke, creating tailored pathways for imaging, thrombectomy, anticoagulation, and intra- and postprocedural monitoring may be beneficial.

Aside from congenital heart disease, many children with stroke have coexisting medical issues such as genetic disorders or infections, which make fluid and blood pressure regulation in the endovascular suite critical. In our experience, close collaboration with pediatric and cardiac intensivists, geneticists, surgeons, and anesthesiologists is mandatory to anticipate issues which may arise throughout the process. Many adult stroke trials excluded patients with baseline moderate disability. While nearly all patients in our series had good functional status at baseline, the infant we treated was critically ill. Whether comorbidities or poor functional baseline should preclude attempts at thrombectomy is not known, and requires further study.

Imaging selection

Optimal imaging criteria for endovascular therapy in children has not been established. DAWN and DEFUSE 3 utilized either MR-DWI or rCT-CBF <30% vols to estimate core infarct size, and MR or CT perfusion parameters (Tmax >6 seconds) to estimate penumbral tissue ‘at risk.’ In DEFUSE 3, a core <70 cc with relatively large mismatch (mismatch ratio >1.8 and volume ≥15 mL) was required to qualify patients for endovascular therapy. Appropriate thresholds for estimating core infarct and critical penumbra in children have not been defined and may differ from adult cutoffs at different ages, since CBF patterns undergo dynamic changes as the brain matures.26 Interestingly, 2/5 patients with perfusion imaging in our series demonstrated a significant volume of Tmax >4, but not >6. Further, these patients had cortical signs on exam that could not be explained by their DWI core, suggesting that Tmax >4 may be a better estimate of critical hypoperfusion for children. ASL is another promising MRI technique to approximate penumbra, which does not require exogenous contrast administration and may be repeated multiple times in the same study. Pitfalls of ASL include its relatively low signal-to-noise ratio and long acquisition time, rendering it susceptible to motion artifact. Studies exploring ASL in adult stroke are conflicting, but some suggest it may overcall true perfusion deficit.27 Optimal imaging techniques and thresholds defining critical hypoperfusion in children have not been defined, and require further study.

While MRI is typically preferred for acute stroke imaging in children,28 it may be contraindicated for certain patients or unavailable at some hospitals. In DAWN and DEFUSE 3, CT perfusion was an alternate imaging option; this technique is feasible in children,29 but is typically avoided due to high radiation burden. Without perfusion, however, estimating core volume and penumbra is challenging. ASPECTS has been explored as another method to estimate core/penumbra with a noncontrast CT,30 and may represent the best option for children and others who cannot undergo MRI or CT perfusion. In our series, patients with CT/CTA alone were considered thrombectomy candidates if their ASPECTS was ≥7 and NIHSS ≥6. Most patients also demonstrated cortical signs (aphasia, neglect) on exam, suggesting a large penumbra.

Our study has a number of limitations. We did not include a control group in our analysis; perfusion patterns in healthy children has been previously explored, but requires further study. The retrospective design introduces recall bias, and the small sample size limits interpretation of these data, but does raise intriguing questions for further study.


Our series suggests that thrombectomy can lead to excellent clinical outcomes in pediatric stroke due to acute large vessel occlusion, and that children who are not treated may have significantly worse outcomes. Perfusion imaging appears feasible in pediatric stroke and may help identify salvageable tissue in extended time windows, though Tmax thresholds defining critical hypoperfusion may differ from adult values. Larger, prospective, multicenter studies are needed to determine optimal clinical and imaging criteria for thrombectomy in this unique population.



  • Contributors All listed authors have fulfilled the following criteria for authorship per ICMJE recommendations: Substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work; AND Drafting the work or revising it critically for important intellectual content; AND Final approval of the version to be published; AND Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. SL conceived the study design, designed the data collection tool, consolidated and analyzed the data, and drafted and revised the paper. GWA conceived the study design, consolidated the data, and significantly revised the paper. MW, NJF, and BJ consolidated and analyzed the patient neuroimaging data and contributed to drafting and revising the paper. EB contributed to data collection tool design and consolidated the data. MM wrote the statistical analysis and cleaned and analyzed the data. JJH, MPM, HMD, and RLD conceived the study design, consolidated the data, enacted and analyzed the neurointerventional procedures and angiography imaging, and revised the paper.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests GWA: Ownership Interest; Significant; iSchemaView. Consultant/Advisory Board; Significant; iSchemaView, Medtronic.

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

  • Patient consent for publication Not required.