Remote Patient Monitoring & Wearables
Teaching AI to detect disease signals in the continuous stream of patient physiology
Industry Challenge
Remote patient monitoring generates continuous streams of physiological data from wearable devices — heart rate, SpO2, respiratory rate, blood glucose, activity levels, and sleep patterns. AI models trained on this data can detect early deterioration, predict exacerbations, and personalise chronic disease management. The annotation challenge is identifying meaningful clinical events in long, noisy time-series data.
How SCILabel Serves This Industry
Data Collection
We source anonymised wearable sensor datasets from digital health research programmes, hospital RPM deployments, and clinical trial data repositories. Data types include continuous ECG, PPG, accelerometry, CGM traces, and multi-parameter vital sign streams.
Data Annotation & Labeling
Our clinical annotators label physiological time-series for clinical events (AF episodes, hypoglycaemic events, sleep apnoea events, fall detection), activity state classification, and anomaly flagging. We also annotate ground truth labels for model training based on corresponding clinical records.
Data & Model Evaluation
Clinical evaluators assess model alert precision and recall against adjudicated clinical event ground truth, test false alert rates, and evaluate model fairness across age, sex, and comorbidity subgroups.
Annotation Types & Formats
- Time-series event annotation: AF, hypoglycaemia, apnoea, falls
- Activity and sleep stage classification on accelerometry
- Anomaly and artefact flagging on continuous physiological streams
- Ground truth label adjudication against clinical records
Specialist Workforce Tracks
Track 2 and Track 4: Biomedical Scientists, Clinical Officers, Nurses, Biomedical Engineers.