- Notifications
You must be signed in to change notification settings - Fork 1.3k
Minutes Data Working Group 20 Apr 2021
Brad edited this page Apr 20, 2021 · 3 revisions
- (Brad) Review the WG-joint work to build a sample FHIR representation of COVID grand challenge data. Suggested datasets:
- COVID-19 2020 Lung CT Lesion Segmentation Challenge - cannot get access
- Other sources: Center for Artificial Intelligence in Medicine & Imaging COVID-19 Data
- (All) Review the discussion at Github on mixed datasets
- (All) Review if there are any Github issues/discussions tagged with #DataWorkingGroup or that should be flagged for this group to discuss
- (Wenqi) Updates from MONAI development
- Attendees: Brad, Raghav, Mona, Stephen
- There should be a library function to extract relevant meta tags from objects
- E.g., DICOM tags; while there is a wealth of data in DICOM tags, depending on the use case, more may be needed
- Discussion of different representations of data
- Re-iterated discussion that FHIR is likely not the right representation right now for training AI models, but may be appropriate for inference
- There needs to be a mapping of different object types into a simplified format for consumption by MONAI, including FHIR -> simplified format
- CSV (or TSV) is too simplified with a lossy destruction of data context (column meaning is human / arbitrarily decided)
- It may make sense to have some form of converter of a format like FHIR or DICOM (or BIDS or etc) into a simplified format
- It's a joint effort between Data and I/O to discuss further
- Parked discussion until Thursday (joint meeting with Data, I/O and other WGs)
- Github Issues and Discussion topics now have the ability to be tagged to a particular WG
- All should review issues and discussions and consider tagging them to Data WG where appropriate
- Based on https://www.kaggle.com/hgunraj/covidxct?select=metadata.csv
- Example is not currently correct (it is a batch of batches; but the child batch elements do not match the correct objects)
Sample FHIR Object:
{ "resourceType": "Bundle", "id": "sequence0", "meta": { "lastUpdated": "2021-04-19T08:00:00-04:00" }, "type": "batch", "entry": [{ "resourceType": "Bundle", "id": "cp_1068", "meta": { "lastUpdated": "2021-04-19T08:00:00-04:00" }, "type": "batch", "entry": [{ "resourceType": "Observation", "id": "endoscope_frame0", "text": { "status": "generated", "div": "<div xmlns=\"http://www.w3.org/1999/xhtml\"><p>(human readable text)</p></div>" }, "status": "final", "category": [{ "coding": [{ "system": "http://terminology.hl7.org/CodeSystem/observation-category", "code": "procedure", "display": "Procedure" }] }], "code": { "coding": [{ "system": "urn:oid:2.16.840.1.113883.6.24", "code": "(appropriate code for endoscope)", "display": "(appropriate label for endoscope)" }] }, "subject": { "reference": "Patient/12345678", "display": "SMITH, J (ID:12345678)" }, "effectiveDateTime": "2020-08-01T08:00:00-04:00", "performer": [{ "reference": "Practitioner/87654321", "display": "DOE, J" }], "device": { "display": "Endoscope" } }, { "resourceType": "Observation", "id": "endoscope_frame0", "text": { "status": "generated", "div": "<div xmlns=\"http://www.w3.org/1999/xhtml\"><p>(human readable text)</p></div>" }, "status": "final", "category": [{ "coding": [{ "system": "http://terminology.hl7.org/CodeSystem/observation-category", "code": "procedure", "display": "Procedure" }] }], "code": { "coding": [{ "system": "urn:oid:2.16.840.1.113883.6.24", "code": "(appropriate code for endoscope)", "display": "(appropriate label for endoscope)" }] }, "subject": { "reference": "Patient/12345678", "display": "SMITH, J (ID:12345678)" }, "effectiveDateTime": "2020-08-01T08:00:00-04:00", "performer": [{ "reference": "Practitioner/87654321", "display": "DOE, J" }], "device": { "display": "Endoscope" } } ] }] }