TIDE is a FHIR R4 Implementation Guide for representing continuous high-resolution time series data (e.g., EEG, ECG, waveform monitoring) using a Metadata-Proxy pattern: structured analytical metadata is carried in FHIR Observation resources, while the underlying raw signal data is referenced externally through TIDEEndpoint resources.
This repository accompanies the manuscript "Exchange of High Frequency Medical Monitoring Data Using Fast Healthcare Interoperability Resources for Interoperable Integration Into Clinical Practice and Research: Methodological Development Study" (submitted to JMIR Medical Informatics) and contains the full source needed to reproduce the FHIR profiles, the evaluation pipeline and the implementation guide referenced there.
| Path | Contents |
|---|---|
TIDE-IG/ |
Buildable FHIR Implementation Guide (SUSHI project: sushi-config.yaml, input/fsh/, input/pagecontent/) — sole source for all profiles, extensions, and terminology |
TIDE-IG/input/examples/ |
Example instances (Patient, Device, Endpoint, Observation) extracted from the evaluation bundle, validated by the IG Publisher |
scripts/ |
Evaluation pipeline: EDF/BIDS ingestion, FHIR resource construction, precision/recall evaluation, synthetic metric validation, bundle export |
scripts/ac2_comparison/ |
Quantitative comparison of TIDE with SampledData, DocumentReference and UV PoCD SampleArrayObservation on an isolated FHIR server |
tide_evaluation_bundle.html |
FHIR transaction Bundle containing all resources produced during the proof-of-concept evaluation, for direct inspection or loading into an independent FHIR R4 server |
| Type | Name | Canonical URL |
|---|---|---|
| Profile | TIDEObservation | http://example.org/tide/StructureDefinition/tide-observation |
| Profile | TIDEObservationChild | http://example.org/tide/StructureDefinition/tide-observation-child |
| Profile | TIDEEndpoint | http://example.org/tide/StructureDefinition/tide-endpoint |
| Extension | RawDataEndpoint | http://example.org/tide/StructureDefinition/tide-rawdata-endpoint |
| Extension | PreviewEndpoint | http://example.org/tide/StructureDefinition/tide-preview-endpoint |
| CodeSystem | TIDECodeSystem | http://example.org/tide/CodeSystem/tide-code-system |
| CodeSystem | TIDESupplementalCodes | http://example.org/tide/CodeSystem/tide-supplemental-codes |
| CodeSystem | TIDEChannelCodeSystem | http://example.org/tide/CodeSystem/tide-channel-code-system |
| CodeSystem | TIDEConnectionTypeCodes | http://example.org/tide/CodeSystem/tide-connection-type-codes |
| ValueSet | TIDEValueSet | http://example.org/tide/ValueSet/tide-value-set |
| ValueSet | TIDEChannelValueSet | http://example.org/tide/ValueSet/tide-channel-value-set |
| SearchParameter | tide-observation-bodysite | http://example.org/tide/SearchParameter/tide-observation-bodysite |
Note: canonical URLs are currently placeholders (
http://example.org/tide/...) and will be replaced with the final canonical namespace upon formal publication of the Implementation Guide.
cd TIDE-IG
sushi build . # FSH -> FHIR JSON (fsh-generated/)
java -jar publisher.jar -ig . -no-sushi # full HTML rendering (output/)
Requires SUSHI and, for the full HTML rendering,
the FHIR IG Publisher
(Java 17+). The build validates all SNOMED CT and LOINC codes against tx.fhir.org; add
-tx n/a only for offline builds, in which case external codes are not checked.
# 1. start Blaze (https://github.com/samply/blaze) on :8080 with the custom bodysite SearchParameter:
docker run -d -p 8080:8080 \
-e DB_SEARCH_PARAM_BUNDLE=/app/tide-search-params.json \
-v "$(pwd)/scripts/tide-search-params.json:/app/tide-search-params.json:ro" \
samply/blaze:1.7.0
pip install mne numpy scipy pandas requests
# 2. place the two OpenNeuro datasets under data/ (see "Data" below), then:
python scripts/tide_pipeline.py # flat model (ds007808)
python scripts/tide_pipeline_v2.py # hierarchical model (ds007823)
python scripts/evaluate.py # precision/recall, example queries, coverage report
# 3. metric validation and bundle export
python scripts/validate_metrics_synthetic.py # ground-truth checks of the metrics on synthetic signals
python scripts/export_bundle.py # exports all resources as a transaction Bundle
python scripts/build_examples.py # splits the bundle into TIDE-IG/input/examples/
# 4. optional: comparison with alternative FHIR approaches (isolated Blaze on :8081)
docker compose -f scripts/ac2_comparison/docker-compose.ac2.yml up -d
python scripts/ac2_comparison/benchmark.py
Outputs are written to results/ (examples for the IG to TIDE-IG/input/examples/).
scripts/tide-search-params.json registers the bodysite SearchParameter
(tide-observation-bodysite) in Blaze. It must be configured before any data is loaded.
Without it, Blaze ignores the unknown bodysite parameter and the channel-level query
returns all Observations instead of the per-channel matches.
This repository does not include raw EEG data. The pipeline uses two public,
CC0-licensed BIDS-EEG datasets from OpenNeuro, which must be downloaded locally
to data/ds007808/ and data/ds007823/ (e.g. with the
OpenNeuro CLI or
DataLad):
| Dataset | Description | DOI |
|---|---|---|
| ds007808 | EEG-Speech Brain Decoding Dataset | 10.18112/openneuro.ds007808.v1.0.0 |
| ds007823 | COVID-19 survivors and close contacts EEG dataset | 10.18112/openneuro.ds007823.v1.0.1 |
The evaluation reported in the manuscript used the following five recordings
(the pipelines process every matching EDF file found under data/, so place only
these files there to reproduce the published bundle):
data/ds007808/sub-03/ses-20240821/eeg/sub-03_ses-20240821_task-speechopen_acq-pangolin_run-01_eeg.edf
data/ds007808/sub-03/ses-20240821/eeg/sub-03_ses-20240821_task-speechopen_acq-pangolin_run-02_eeg.edf
data/ds007823/sub-CUCOV003/eeg/sub-CUCOV003_task-COVID_eeg.edf
data/ds007823/sub-CUCOV008/eeg/sub-CUCOV008_task-COVID_eeg.edf
data/ds007823/sub-CUCOV020/eeg/sub-CUCOV020_task-COVID_eeg.edf
Each pipeline also needs the BIDS sidecar files next to each EDF file (*_eeg.json,
*_channels.tsv, *_events.tsv).
The resulting FHIR resources do not duplicate the raw signal files: TIDE's
TIDEEndpoint resources reference the original EDF files on OpenNeuro
(connection type direct-https from TIDEConnectionTypeCodes).
Corresponding author:
Dr. René Hosch
University Hospital Essen
Hufelandstraße 55, 45147 Essen
rene.hosch@uk-essen.de
Yutong Wen M. Sc.
University Hospital Essen
Hufelandstraße 55, 45147 Essen
yutong.wen@uk-essen.de
Sara Erma Kaya M. Sc.
University Hospital Essen
Hufelandstraße 55, 45147 Essen
sara.kaya@uk-essen.de