A Two-Stream Graph Convolutional Network Based on Brain Connectivity for Anesthetized States Analysis

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2022, Volume and Issue: 30, P. 2077 - 2087

Published: Jan. 1, 2022

Investigating neural mechanisms of anesthesia process and developing efficient anesthetized state detection methods are especially on high demand for clinical consciousness monitoring. Traditional monitoring not involved with the topological changes between electrodes covering prefrontal-parietal cortices, by investigating electrocorticography (ECoG). To fill this gap, a framework based two-stream graph convolutional network (GCN) was proposed, i.e., one stream extracting structure features, other node features. The includes GCN Model 1 2. For 1, brain connectivity networks were constructed using phase lag index (PLI), representing different A common adjacency matrix founded through dual-graph method, features expressed nodes. Therefore, traditional spectral can be directly applied graphs changing structures. On hand, average absolute signal amplitudes calculated as then fully connected these input This method learns both nodes graph, uses approach to enhance focus Based ECoG signals monkeys, results show that which distinguish awake state, moderate sedation deep achieved an accuracy 92.75% in group-level experiments mean 93.50% subject-level experiments. Our work verifies excellence monitoring, recognition also shows may carry neurological markers associated anesthesia.

Language: Английский

---

Generating ten BCI commands using four simple motor imageries and classification by divergence-based DNN

Neural Computing and Applications, Journal Year: 2022, Volume and Issue: 35(2), P. 1303 - 1322

Published: Sept. 26, 2022

Language: Английский

---

Wading corvus optimization based text generation using deep CNN and BiLSTM classifiers

Biomedical Signal Processing and Control, Journal Year: 2022, Volume and Issue: 78, P. 103969 - 103969

Published: Aug. 5, 2022

Language: Английский

---

Mixed Dimension Convolutional Neural Network for Bispectral Index Estimation

Published: Jan. 1, 2024

Bispectral index (BIS), which ranges from 0 to 100 degrees, is a more precise approach using electroencephalography (EEG) evaluate the level of sedation and monitor depth anesthesia. However, it challenging correctly instantly estimate BIS for anesthesia surgery due algorithms' secrecy, postoperative complications, unanticipated consciousness. We have suggested estimation based on deep learning as regression model address this issue. The was trained large dataset consecutive brain-electrical signals 60s. Therefore, paper introduces an in network that combines mixed dimension features time domain frequency increase accuracy results. During training, by introducing data obtained Fast Fourier Transform, improving model, adopting two-stream convolution structure, characteristics operations simultaneously optimize image recognition performance. Despite unbalanced distribution forecasts, experimental results demonstrated our structure produced convincing root mean square error (3.79±1.8) absolute (1.93±0.61), outperforming other approaches currently use.

Language: Английский

---

Msrtnet: Multi-Scale Spatial Residual Network Based on Time-Domain Transformer

Published: Jan. 1, 2024

Language: Английский

---