A Deep Learning Approach to Grading Neonatal Hypoxic-Ischemic Encephalopathy Using ECG Spectrograms

The grade of Hypoxic-Ischemic Encephalopathy (HIE), a condition caused by cerebral oxygen deprivation, can be determined through analysis of Electroencephalogram (EEG). Neonates with moderate to severe HIE can be treated with therapeutic hypothermia, which has prompted the development of an automatic HIE grading system. Electrocardiogram (ECG) signals are easier and more accessible to obtain than the EEG in the Neonatal Unit and may help grade HIE. This research explores two different approaches for grading HIE, leveraging a large ECG dataset from newborns. The classical approach involves calculating heart rate (HR) from ECG signal followed by feature extraction in both time and frequency domains and classification using Random Forest and multilayer perceptron. In contrast, the proposed approach extracts the spectrograms directly from the ECG signal to serve as input for a convolutional neural network. The results indicate that the proposed method achieves a higher AUC while bypassing the time-consuming process of HR calculation and enabling the use of more robust deep learning models.Clinical relevance- Utilizing newborn ECG signals as an alternative to EEG signals for automatic HIE grading may provide a more accessible method for clinicians to grade encephalopathy and evaluate the need for hypothermia treatment.