filipelopesmedbr/CIEL-Clinical-Concepts-to-ICD-11

TD;LR Run:

pip install torch==2.4.1+cu118 torchvision==0.19.2+cu118 torchaudio==2.4.1 --extra-index-url https://download.pytorch.org/whl/cu118
pip install -U packaging setuptools wheel ninja
pip install --no-build-isolation axolotl[flash-attn,deepspeed]
axolotl train axolotl_2_a40_runpod_config.yaml

📚 CIEL to ICD-11 Fine-tuning Dataset

This dataset was created to support the fine-tuning of open-source large language models (LLMs) specialized in ICD-11 terminology mapping.

It includes clinical concepts from the Columbia International eHealth Laboratory (CIEL) and official entities from the ICD-11 MMS classification by the World Health Organization (WHO).

⸻

🍿 Dataset Structure

Each line in the .jsonl file follows this structure:

{
  "instruction": "Provide the ICD-11 code for the clinical concept. Include extensions or cluster codes if required for full specificity.",
  "input": "Carcinoma of tongue base - dorsal surface",
  "output": "SAME-AS 2B40.0&XA7LA4",
  "category": "source:ciel>same-as+stem+extensions"
}

• —instruction: Text prompting the model to generate the expected output.
• —input: Clinical concept to be mapped.
• —output: Expected mapping output with the relation type (e.g., SAME-AS, NARROWER-THAN, BROADER-THAN) followed by the ICD-11 code(s).
• —category: Metadata field indicating the source and mapping type.

⸻

📂 Categories

The dataset covers multiple mapping scenarios:

⸻

🎯 Intended Use

• —Fine-tuning LLMs to understand, map, and post-coordinate ICD-11 concepts.
• —Training specialized RAG (Retrieval Augmented Generation) systems for healthcare terminology.
• —Supporting automated ICD-11 coding tasks.

⸻

⚡ Notes

• —Clinical concepts were carefully filtered and reviewed to ensure quality.
• —Mapping relationships (e.g., SAME-AS, NARROWER-THAN) follow standardized conventions.
• —Extension codes (X...) are never returned alone — they must be attached to a stem or cluster.

⸻

🧪 License

This dataset is distributed under the MIT License.

⸻

🤝 Contributions

Contributions, corrections, and suggestions are welcome to improve this resource for the community.