Research

Biosignals

Biosignal Analysis and Neurological Applications

Biosignals are continuous, noisy, and clinically valuable. The program builds models that generalize across datasets and support expert-level interpretation.

01EEG and physiological signals
02Signal representation
03Transformer or neural classifier
04Sleep, seizure, and monitoring insight

Recent papers

What this program is building

Selected recent and foundational papers, summarized around the task, why it matters, and the main technical result.

2025AAAI

Long-Term EEG Partitioning for Seizure Onset Detection

1Long EEG stream
2Temporal partitioning
3Seizure onset alert
Task
Detect seizure onset from long-duration EEG by partitioning signals into useful temporal regions.
Why it matters
Long EEG streams are difficult to review manually; better partitioning can help clinicians focus attention.
Main result
The method targets long-horizon EEG structure so seizure onset detection can operate beyond short windows.
Paper details
2024NeurIPS

BIOT: Biosignal Transformer for Cross-data Learning in the Wild

1Mixed biosignals
2Biosignal transformer
3Cross-dataset prediction
Task
Train a biosignal transformer that generalizes across heterogeneous datasets and recording settings.
Why it matters
Real biosignal data vary by device, montage, cohort, and protocol, so cross-data learning is essential for deployment.
Main result
BIOT provides a transformer architecture for learning robust representations across biosignal datasets.
Paper details
2024Sleep

What Radio Waves Tell Us about Sleep!

1Radio sensing
2Sleep signal inference
3Contactless monitoring
Task
Use radio-wave sensing to infer sleep-related signals without conventional attached sensors.
Why it matters
Contactless monitoring can make sleep assessment more comfortable and scalable outside the lab.
Main result
The work explores how ambient sensing can contribute to sleep understanding and monitoring.
Paper details
2023Neurology

Development of Expert-Level Classification of Seizures and Rhythmic and Periodic Patterns During EEG Interpretation

1EEG recording
2Pattern classifier
3Expert-level labels
Task
Classify seizures and rhythmic/periodic EEG patterns at expert-level performance.
Why it matters
EEG interpretation is specialized and time-intensive; reliable AI can support expert review.
Main result
The study demonstrates AI assistance for clinically meaningful EEG pattern classification.
Paper details
2019MLHC

EEGtoText: Learning to Write Medical Reports from EEG Recordings

1EEG signal
2Signal-to-text model
3Draft report
Task
Generate report-style text from EEG recordings.
Why it matters
It connects signal modeling with clinical communication, a key step for usable AI systems.
Main result
The paper studies how EEG representations can be translated into clinically readable report language.
Paper details
2018JAMIA

Expert-Level Sleep Scoring with Deep Neural Networks

1Sleep signals
2Deep sleep scorer
3Sleep-stage timeline
Task
Automatically score sleep stages from physiological recordings.
Why it matters
Sleep scoring is laborious and clinically important; automation can support scalable sleep medicine.
Main result
Deep neural networks reach expert-level sleep staging performance in this foundational Sunlab line of work.
Paper details

Representative publication links

NeurIPS · 2024

BIOT: Biosignal Transformer for Cross-data Learning in the Wild

Transformer architecture for cross-data biosignal learning.

Paper details Neurology · 2023

Development of Expert-level Classification of Seizures and Rhythmic Patterns during EEG Interpretation

Expert-level EEG interpretation support.

Paper details JAMIA · 2018

Expert-level Sleep Scoring with Deep Neural Networks

Deep learning for sleep staging.

Paper details

Interested in this program?

Send a concise note with the program name, your role, the problem you want to work on, and any relevant data, code, clinical setting, or research experience.

Contact Sunlab
Clinical AI Medical LLMs Drug Discovery