Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →mkurman/medcat
Medical Concept Annotation Toolkit. Trainable NLP for extracting clinical concepts from unstructured text. Supports ICD-10, SNOMED CT, RxNorm, UMLS. Active learning for custom medical ontologies.
Passed 18 of 22 test cases with the skill, and 13 of 22 without it.
and 3 more in the full report
5 fixed13 passed either way4 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
What is the fourth step in the MedCAT workflow after mapping concepts, used to improve model performance on edge cases?
| Checked | Without skill | With skill |
|---|---|---|
| Identifies active learning training to correct errors and add concepts | Missed | Passed |
The instructions your model reads when the skill loads.
MedCAT trains NLP models for extracting clinical concepts from unstructured text. Supports ICD-10, SNOMED CT, RxNorm, UMLS, and custom ontologies with active learning.
uv pip install medcat
from medcat.cat import CAT
cat = CAT.load_model_pack("medcat_model_pack.dat")
text = "Patient with type 2 diabetes and hypertension, prescribed metformin 500mg BID."
doc = cat(text)
for entity in doc.entities:
print(f"{entity.name:<25} {entity.cui:<10} confidence={entity.confidence:.2f}")
# type 2 diabetes D003920 confidence=0.97
# hypertension D006973 confidence=0.99
cat.add_cui_to_category("D003920", "Diabetes Mellitus")
cat.train(text="Patient has diabetes", cui="D003920", value="Diabetes Mellitus")
unmatched = cat.get_unmatched_concepts() # concepts needing review
test cases passed, vs 13 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.