Overview
MEDWEAR is validated through two ongoing pilot studies in collaboration with clinical and robotics research groups at ETH Zurich. Both pilots share the same end destination - the MEDWEAR data collection platform - but differ in how data flows from the wearable to that platform: one is offline via a smartphone app, the other is online via ROS 2 middleware.
SPZ - Offline data collection
Context
Clinical study in collaboration with SPZ (Schweizerisches Paraplegiker-Zentrum). Participants wear sensors during daily activities outside a lab environment. The study focuses on capturing physiological signals - PPG, EDA, and IMU - using consumer-grade wearable devices.
Data flow
- A wearable device streams PPG, EDA, and IMU signals.
- A smartphone running the SCAI app collects and buffers the raw data locally.
- Batches are uploaded to an intermediate server for secure storage.
- A post-processing pipeline converts the data to MEDWEAR format (JSON Schema validated).
- Standardized data is ingested into the MEDWEAR platform for visualization, access control, and export.
Why offline?
Participants may not always have reliable internet connectivity. The smartphone buffers data locally and syncs when a connection is available, decoupling data collection from transmission.
Signals
LLUI - Online cognitive load estimation
Context
Research study at ETH Zurich's LLUI (Lower-Limb User Interaction) lab. Participants wear sensors during human-robot interaction tasks. The goal is real-time cognitive load estimation to adapt assistive robot behavior based on the user's physiological state.
Data flow
- A wearable device streams PPG, ECG, EDA, and IMU signals continuously.
- Raw data is immediately formatted as MEDWEAR messages at the source.
- ROS 2 middleware transports the data in real time using
healthcare_msgsmessage definitions. - The MED-API ROS 2 bridge ingests the stream into the platform with low latency.
- The MEDWEAR platform serves the data for live dashboards, role-based access, and downstream ML pipelines.
Why online?
Cognitive load estimation requires sub-second latency. An online pipeline eliminates buffering delays and enables closed-loop feedback between the wearable signals and the robot controller.
Signals
What the Pilots Validate
| Property | SPZ (Offline) | LLUI (Online) |
|---|---|---|
| Connectivity model | Asynchronous batch upload | Synchronous real-time stream |
| Transport layer | SCAI smartphone app | ROS 2 + MED-API bridge |
| Schema format | JSON Schema (post-hoc conversion) | ROS 2 healthcare_msgs (native) |
| Primary signal types | PPG, EDA, IMU | PPG, ECG, EDA, IMU |
| Primary use | Clinical data collection & export | Real-time cognitive load estimation |
| Downstream consumer | Researcher / clinician dashboard | Closed-loop robot controller + dashboard |