Value of cerebral microhemorrhages detected with susceptibility-weighted MR Imaging for prediction of long-term outcome in children with nonaccidental trauma

Radiology. 2010 Sep;256(3):898-905. doi: 10.1148/radiol.10091842.


Purpose: To determine the prevalence of parenchymal brain microhemorrhages (MHs) in infants with nonaccidental trauma (NAT) by using susceptibility-weighted (SW) magnetic resonance (MR) imaging and to assess whether the presence of MH results in improved prediction of the long-term neurologic outcome.

Materials and methods: A retrospective case-control analysis of the data for 101 children aged 1-32 months with forensic pediatric specialist-confirmed NAT was performed with institutional review board approval. Sixty-two patients were boys (mean age, 8.4 months +/- 7.4 [standard deviation]), and 39 were girls (mean age, 7.4 months +/- 7.8). The imaging findings and clinical data of the children who were examined with SW imaging were collected. Exclusion criteria included pre-existing cognitive delays, central nervous system malformations, previous brain injuries, and/or birth before 30 weeks gestation. Dichotomized long-term neurologic outcomes (good [normal, mild disability, or moderate disability] versus poor [severe disability, vegetative state, or death]) at greater than or equal to 6 months (mean, 33 months; range 6-95 months) were available for 53 patients (36 boys [mean age, 7.3 months +/- 5.9]; 17 girls [mean age, 7.4 months +/- 7.9]; overall range, 2-32 months). Logistic regression was used to determine whether the presence of SW imaging-depicted MH, as compared with other radiologic findings, resulted in improved prediction of long-term neurologic outcome.

Results: Imaging findings showed that of the 101 patients, 29 (29%) had MH at SW imaging, 66 (65%) had extraaxial hemorrhages, 52 (51%) had retinal hemorrhages, and 35 (35%) had evidence of acute ischemic injury. A significantly larger number of children with poor outcomes than children with good outcomes had brain MH (nine of 14 vs seven of 39; P = .001) and ischemic injury (13 of 14 vs 17 of 39; P = .006). Logistic regression analysis revealed presence of MH at SW imaging-followed by acute ischemic injury, initial Glasgow Coma Scale score, and age-to be the most significant single variable in the final model, with an overall predictive accuracy of 92.5%.

Conclusion: Presence of intraparenchymal brain MH in children with NAT, as detected on SW images, correlates with significantly poor long-term neurologic outcome, improves outcome prediction compared with the predictions made by using other tested clinical and imaging findings, and is most predictive when combined with presence of ischemic injury.

MeSH terms

  • Case-Control Studies
  • Cerebral Hemorrhage / diagnosis*
  • Cerebral Hemorrhage / pathology
  • Child Abuse / diagnosis*
  • Female
  • Glasgow Coma Scale
  • Humans
  • Infant
  • Logistic Models
  • Magnetic Resonance Imaging / methods*
  • Male
  • Predictive Value of Tests
  • Prognosis
  • Retrospective Studies
  • Statistics, Nonparametric