curl --request POST \
--url https://api.eka.care/med-link/nel/link \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"query": "chest pain",
"ontology": "snomed",
"version": "20250401_extended",
"top_k": 3
}
'{
"request_id": "0e95448d-7e32-44c9-8586-c33315412330",
"query": "chest pain",
"query_breakdown": null,
"ontology": "snomed",
"version": "20250401_extended",
"index_name": "snomed_20250401_extended_c88b15811956d4fcb54da9ae_1bfcc90540_pg",
"model_version": "c88b15811956d4fcb54da9ae6d14a733ac90e7a7",
"results": [
{
"term_id": "29857009",
"term_name": "Chest pain",
"score": 1,
"is_linked": true,
"metadata": {
"semantic_tag": "finding",
"is_core": true,
"source": "snomed_ct_20250401_extended",
"language": "en",
"script": "latin"
}
},
{
"term_id": "29857009",
"term_name": "Chest pain (finding)",
"score": 0.9494552299218054,
"is_linked": true,
"metadata": {
"semantic_tag": "finding",
"is_core": true,
"source": "snomed_ct_20250401_extended",
"language": "en",
"script": "latin"
}
}
]
}Medical Entity Codification
Link Entity
Links a single free-text medical entity to candidate codes from the requested ontology. The metadata field carries ontology-specific hints — it is strongly recommended for LOINC (a unit can change the linked code) and optional for medication. SNOMED CT and ICD-10-CM Comprehend do not use metadata.
POST
/
med-link
/
nel
/
link
curl --request POST \
--url https://api.eka.care/med-link/nel/link \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"query": "chest pain",
"ontology": "snomed",
"version": "20250401_extended",
"top_k": 3
}
'{
"request_id": "0e95448d-7e32-44c9-8586-c33315412330",
"query": "chest pain",
"query_breakdown": null,
"ontology": "snomed",
"version": "20250401_extended",
"index_name": "snomed_20250401_extended_c88b15811956d4fcb54da9ae_1bfcc90540_pg",
"model_version": "c88b15811956d4fcb54da9ae6d14a733ac90e7a7",
"results": [
{
"term_id": "29857009",
"term_name": "Chest pain",
"score": 1,
"is_linked": true,
"metadata": {
"semantic_tag": "finding",
"is_core": true,
"source": "snomed_ct_20250401_extended",
"language": "en",
"script": "latin"
}
},
{
"term_id": "29857009",
"term_name": "Chest pain (finding)",
"score": 0.9494552299218054,
"is_linked": true,
"metadata": {
"semantic_tag": "finding",
"is_core": true,
"source": "snomed_ct_20250401_extended",
"language": "en",
"script": "latin"
}
}
]
}About this endpoint
POST /med-link/nel/link links a single free-text medical entity to candidate
codes from the requested ontology. Pick the ontology and version from
List Registry.
Use the example selector above to see a request and response for each ontology.
The metadata field by ontology
metadata carries ontology-specific hints. It changes results for LOINC and
medication, and is ignored by SNOMED CT and ICD-10-CM Comprehend.
SNOMED CT — snomed
No metadata required. Send the symptom, finding or disorder as the query.
Each result’s metadata includes a semantic_tag (e.g. finding, disorder).
LOINC — loinc
metadata is strongly recommended — the lab test name alone is often
ambiguous, and a unit or specimen decides which LOINC code is correct.
| Key | Type | Description |
|---|---|---|
unit | string | Test unit, e.g. g/dl, mg/dl, %. Most important hint. |
value | string | Numeric result value. Improves value-type matching. |
range | string | Reference range for the result. |
specimen | string | Specimen type, e.g. blood, urine. |
specimen_text | string | Free-text specimen description (auto-detected if omitted). |
panel_name | string | Panel name for multi-test panels. |
extended_test_name | string | Extended/alternate test name. |
Without
metadata, a LOINC query such as hemoglobin can link to the wrong
specimen (e.g. urine instead of blood). Supplying unit and specimen lets the
pipeline return is_linked: true against the correct code.Medication — medication
metadata is optional — it helps resolve ambiguous brand names. The score
field is null; read metadata.matching_breakdown and metadata.linking_requirements
on each result instead.
| Key | Type | Description |
|---|---|---|
form | string | Medication form, e.g. tablet, capsule, injection. |
generic_name | string | Generic / salt name of the drug. |
dose_unit | string | Dosage unit, e.g. mg, mcg, ml. |
dose_custom | string | Custom dosage string. |
ICD-10-CM Comprehend — icd-10-cm (comprehend)
No metadata required. This version uses AWS Comprehend Medical, so it works
well on natural-language clinical text. The response includes a query_breakdown
with the detected entity, traits and attributes.
Reading the response
| Field | Description |
|---|---|
results[] | Candidate codes, ranked by relevance. |
results[].term_id | The ontology code. |
results[].is_linked | true when the pipeline is confident of a single correct match. |
results[].score | Relevance/confidence (null for medication). |
query_breakdown | How the query was parsed (LOINC, medication, ICD-10-CM Comprehend; null for SNOMED CT). |
Authorizations
Bearer access token issued by the Eka authorization flow. See Authorization.
Body
application/json
Response
Candidate codes for the query, ranked by relevance.
Was this page helpful?

