Toward convenient depression detection using two non-hair-bearing frontal EEG channels

Depressive disorder (DD) is a prevalent mental illness with substantial impacts on emotional, cognitive, and physical functioning. The diagnosis of DD however is heavily relying on subjective assessments. Although some exploratory works proposed objective detection methods based upon EEG signals, the widely-used high-density setup significantly limits its practicability and convenience.

Here, we introduced an analysis framework for convenient DD detection using EEG data recorded at two non-hair-bearing (NHB) frontal channels (Fp1 and Fp2) from 80 participants (controls/patients = 38/42). Multiscale coarse-graining strategies were then employed on the EEG data to construct scale-dependent temporal representations, where EEG features were extracted on such reconstructed temporal sequences and set as input for the following classification models. The impact of various settings (i.e., window length of EEG data, coarse-graining designs, and scale factors) was also quantitatively investigated.

We achieved satisfactory performance (> 88%) for DD classification using NHB EEG signals. Further analyses showed that moderately long window (> 12s) providing more stable performance. Moreover, overlapping coarse-graining strategies exhibited superior robustness in comparison with non-overlap strategy. Incorporating multiscale temporal representations consistently improved discriminative capability.

These findings indicate that under frontal NHB EEG configurations, multiscale temporal organization is not merely an optional enhancement but a necessary representational strategy for reliable DD detection. The proposed framework provides practical guidance for developing efficient, wearable EEG-based systems for depression screening in real-world settings.