LIVE TELEMETRY
Research Dashboard
Monitor the connected COPD-EFF model, dataset readiness, and research telemetry from the respiratory acoustic pipeline.
ServiceDemo / offline
Audio files—
Test accuracy—
ModelCOPD-EFF
Connected track
COPD-EFF / binary screening
EfficientNetB0 + BiLSTM + 4-head attention. The live service returns COPD or NON-COPD probabilities.
Sensitivity—
Specificity—
F1 score—
ROC-AUC—
Dataset readiness
Check the evidence before you read the metrics.
Run the connected COPD-EFF dataset validator to inspect audio files, annotations, patient IDs, and split readiness.
Illustrative training trace
Replace with persisted training history when available
Class Distribution
Illustrative reference distribution · COPD-EFF is binary
Wheeze (40%)
Crackle (30%)
Normal (30%)
Recent Inferences
Illustrative local history until persistence is connected
| ID | File | Prediction | Confidence | Time |
|---|---|---|---|---|
| INF-001 | patient_A_342.wav | Wheeze | 92% | 2 mins ago |
| INF-002 | patient_B_112.wav | Normal | 96% | 15 mins ago |
| INF-003 | patient_C_089.wav | Crackle | 84% | 1 hour ago |
| INF-004 | patient_D_551.wav | Mixed | 76% | 3 hours ago |