In Brain and behavior
OBJECTIVE : Prematurity is associated with a high risk of long-term behavioral problems. This study aimed to assess the prognostic utility of volumetric brain data at term-equivalent-age (TEA), clinical perinatal factors, and parental social economic risk in the prediction of the behavioral outcome at 5 years in a cohort of very preterm infants (VPT, <32 gestational weeks).
METHODS : T2-weighted magnetic resonance brain images of 80 VPT children were acquired at TEA and automatically segmented into cortical gray matter, deep subcortical gray matter, white matter (WM), cerebellum (CB), and cerebrospinal fluid. The gray matter structure of the amygdala was manually segmented. Children were examined at 5 years of age with a behavioral assessment, using the strengths and difficulties questionnaire (SDQ). The utility of brain volumes at TEA, perinatal factors, and social economic risk for the prediction of behavioral outcome was investigated using support vector machine classifiers and permutation feature importance.
RESULTS : The predictive modeling of the volumetric data showed that WM, amygdala, and CB volumes were the best predictors of the SDQ emotional symptoms score. Among the perinatal factors, sex, sepsis, and bronchopulmonary dysplasia were the best predictors of the hyperactivity/inattention score. When combining the social economic risk with volumetric and perinatal factors, we were able to accurately predict the emotional symptoms score. Finally, social economic risk was positively correlated with the scores of conduct problems and peer problems.
CONCLUSIONS : This study provides information on the relation between brain structure at TEA and clinical perinatal factors with behavioral outcome at age 5 years in VPT children. Nevertheless, the overall predictive power of our models is relatively modest, and further research is needed to identify factors associated with subsequent behavioral problems in this population.
Liverani Maria Chiara, Loukas Serafeim, Gui Laura, Pittet Marie-Pascale, Pereira Maricé, Truttmann Anita C, Brunner Pauline, Bickle-Graz Myriam, Hüppi Petra S, Meskaldji Djalel-Eddine, Borradori-Tolsa Cristina
2023-Jan-14
MRI, behavioral outcome, classification, machine learning, preterm infants, volumetric brain data
