Research project
Multimodal Sleep Apnea Screening
A research pipeline that combines respiratory acoustics and physiological dynamics for OSA screening.
Problem
Conventional diagnosis is resource-intensive; research explores whether smaller, lower-burden signal sets can retain clinically useful information.
Why it matters
The long-term goal is accessible, interpretable and robust screening that can guide who should receive full diagnostic testing.
Pipeline
The pipeline remains readable without animation and progressively activates while scrolling.
01
Signals
Acquire respiratory and cardiovascular modalities
SnoreAirflowECGPPG
02
Preprocessing
Align, resample, window and normalise
03
Encoders
Learn modality-specific temporal representations
04
Fusion
Combine complementary information
05
Screening output
Estimate respiratory-event / OSA-related class
Signals
Snore intensityAirflowECGPPGSpO₂
Models
ResNet-1DCNN-LSTMFusion networks