Application of deep learning algorithms to neural function assessment following cardiac arrest

  • Accurate assessment of neural function for individuals in a coma following cardiac arrest is challenging and currently relies on subjective scoring of physiological signals.
  • Using convolutional neural networks to model EEG responses to standardized auditory stimuli demonstrated positive predictive power when predicting awakening for both individuals undergoing therapeutic hyperthermia and normothermia (0.83 ±0.04 and 0.81 ±0.05, respectively).
  • The authors concluded that deep learning algorithms such as convolutional neural networks, in combination with auditory stimulation, have potential to standardize neural function assessment and likelihood of awakening from coma.