Predicting depth of anaesthesia from single-channel EEG using a deep TCN-BiLSTM-attention model with EWMA
Sukriti, Chirag Kriplani, Suman Kumar, Abhishek Singh · 2026
Read Predicting depth of anaesthesia from single-channel EEG using a deep TCN-BiLSTM-attention model with EWMA by Sukriti, Chirag Kriplani, Suman Kumar, Abhishek Singh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13473519.txt.
Description
Sukriti and three colleagues built in 2026 a transparent system for tracking anaesthesia depth from a single EEG lead. Getting depth right protects patients and allows tailored dosing, but the widely used Bispectral Index monitor relies on proprietary algorithms that clinicians cannot inspect. Their framework estimates the index continuously from raw single-lead brain waves, using a temporal convolutional network, a compact bidirectional LSTM, and attention pooling, smoothed with an exponentially weighted moving average.
On a random segment split of a public perioperative dataset, the model reached a mean absolute error of 4.499 and a correlation of 0.903 with reference values. Under a stricter test on unseen patients, error rose as expected, to a mean absolute error of 6.03, with concordance of 0.819, yet results stayed stable across folds. With about 1.07 million parameters and roughly 430 segments processed per second, the design suits near-real-time bedside use, offering an open alternative to black-box monitors.
Catalog record
| CREATOR | Sukriti, Chirag Kriplani, Suman Kumar, Abhishek Singh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |