Abstracts

Rationale: Neonates with acute symptomatic seizures can have a range of long-term developmental outcomes, ranging from normal development to profound impairment. Electroencephalographic (EEG) recordings provide a direct assay of neuronal activity with a range of age-specific patterns that coincide with normal development, and severely abnormal early findings on EEG increase the risk for abnormal long-term neurodevelopmental outcomes. Whether the absence of specific developmental graphoelements could provide long-term prognostic information is unknown. A candidate graphoelement are sleep spindles: discrete bursts of 9-16 Hz oscillations during non-rapid eye movement (NREM) sleep that are generated by the reticular nucleus of the thalamus. Spindles typically present by 6 weeks of age and are consistently observed during NREM sleep after the 9th postnatal week. Spindle activity reflects thalamocortical circuit function, where spindle deficits reflect dysfunction of this circuit and expected impact on neurological function. We therefore hypothesized that decreased spindle activity in the Rolandic regions in young infants would predict long-term contralateral motor impairment.

Methods: EEGs from 17 infants with a history of acute symptomatic neonatal seizures, the ‘at risk’ group (7F, ages 2.55±0.99 months), and 24 control infants who had normal neurodevelopment, non-epileptic events, and normal EEG (14F, ages 2.20±0.81 months) were analyzed. All EEGs were manually reviewed, and all available NREM stages 2 and 3 data were included for analysis. For the ‘at risk’ group, motor outcome was obtained from chart review (ages 3.99±1.46 years), where any noted weakness or increased tone in the upper or lower extremities were included as poor motor outcome. To quantify spindles, we validated an automated infant spindle detector using a latent state model (LSM). To do so, two reviewers hand-marked 3362 sleep spindles from EEGs in 47 healthy infants ages 0-24 months, by consensus. The LSM detector was then trained and validated using leave-one-out cross-validation. The validated detector was then applied to the ‘at risk’ and control infant EEG NREM datasets. Spindle features were computed in the C3 and C4 channels and compared to motor outcome using logistic regression.

Results: The LSM infant spindle detector had excellent performance compared to manual markings (F1=0.50). Including both the ‘at risk’ infants and controls, spindle rate, duration, and percentage were decreased in the hemispheres corresponding to poor contralateral motor outcome (p=0.003, p=0.004, p=0.001, respectively). These differences were also evident among the ‘at risk’ infants only (p=0.002, p=0.015, p=0.011, respectively), showing decreased activity of spindles in the contralateral Rolandic regions among the infants with long-term motor impairment (Figure 1A-C).

Conclusions: Decreased spindle activity reflects disrupted thalamocortical circuit function and provides an early prognostic indicator of poor motor outcome in infants. Future work will evaluate whether spindle features can predict the degree of motor dysfunction in a larger dataset.

Funding: Please list any funding that was received in support of this abstract.: N/A.