Extraction and analysis of abnormal EEG features in children with amblyopia

Early and accurate diagnosis of amblyopia is crucial for the healthy development of children. Existing clinical diagnostic methods rely on patient cooperation, which can easily lead to misdiagnosis. The commonly used features derived from visual evoked potentials (VEP) only provided limited information for characterizing the whole brain, highlighting the need for integrating additional data sources, such as brain network metrics, to achieve a more comprehensive understanding.

We extracted 488 features from 64-channel EEG data recorded from thirty amblyopic children. The features mainly derived from a weighted functional brain network based on coherence across different frequency bands. Feature selection and linear classification techniques were employed to assess their effectiveness in distinguishing amblyopia from normal children.

Abnormal EEG features were distributed not only in the occipital lobe but also in non-visual regions, with a higher prevalence in the alpha and beta bands. Their decoding performance surpassed traditional VEP features, and their combination achieved the highest accuracy (89.00%). Moreover, features beyond the occipital lobe exhibited limited decoding performance when considered individually, yet they still have an obvious contribution.

The study identified novel abnormal EEG features associated with amblyopia and demonstrated the potential of multi-channel EEG recordings to assist in the diagnosis of amblyopia.

The study suggests amblyopia may impair more abilities beyond visual cognition and further provides objective biomarkers for diagnosing amblyopia, which is essential for effective treatment.