Skip to main content
. 2022 Mar 4;9:822810. doi: 10.3389/fmolb.2022.822810

FIGURE 1.

FIGURE 1

The proposed Transformer-based HFO detection framework is designed specifically for the presurgical diagnosis of biomedical one-dimensional MEG signal data. Briefly, the HFO classification framework includes signal segmentation, signal augmentation, TransHFO signal classification, and signal labeling. This framework achieves more robust and reliable performance on HFO classification than baseline models. Furthermore, we find that shallow TransHFO ( < 10 layers) outperforms deep TransHFO (≥10 layers) on most data augmented factors, revealing the importance of human labeled data and the potential of deep-learning methods for automatic diagnosis of medical signal.