Datasets:

AmareshHebbar
/

radiology-coder-sft

license: apache-2.0
language:
  - en
tags:
  - medical
  - radiology
  - icd-10
  - report-coding
  - clinical-nlp
  - imaging
task_categories:
  - text-generation
pretty_name: Radiology Report ICD-10 Coder
size_categories:
  - 10K<n<100K

Radiology Report ICD-10 Coder

Part of the AxisMapper Medical AI Suite — 16 domain-specific SFT datasets for fine-tuning medical LLMs.

Built by AmareshHebbar | Studio Ilios / Humanova Minds

What this dataset does

Radiology reports / impressions → ICD-10-CM codes for all documented findings

Why download this

Automate radiology coding, build report-to-code pipelines for RIS/PACS integration, or train models to extract diagnosis codes from chest X-ray, CT, MRI, and ultrasound impressions.

Dataset stats

Split	Rows
Train	20,072
Validation	2,509
Test	2,509
Total	25,090

Data format

Every row is a messages list in chat format — compatible with Unsloth, TRL SFTTrainer, LLaMA-Factory, and any OpenAI-style fine-tuning pipeline:

{
  "messages": [
    {"role": "system",    "content": "You are a ..."},
    {"role": "user",      "content": "IMPRESSION: 1.8cm hypoechoic nodule right thyroid lobe, TIRADS 4. Recommend FNA."},
    {"role": "assistant", "content": "ICD-10-CM Coding: E04.1 — Non-toxic single thyroid nodule
Note: Add malignancy code post-biopsy if confirmed."}
  ]
}

Data source

AGBonnet/augmented-clinical-notes — 25k radiology-relevant clinical notes filtered by imaging keywords → https://huggingface.co/datasets/AGBonnet/augmented-clinical-notes

All data is extracted from authoritative public sources. No LLM-generated or synthetic content.

Who should use this

Radiology AI companies, RIS/PACS vendors, hospital coding departments, teleradiology platforms.

Quick start

from datasets import load_dataset

ds = load_dataset("AmareshHebbar/radiology-coder-sft")
print(ds["train"][0])

Fine-tuning example (Unsloth)

from unsloth import FastLanguageModel
from trl import SFTTrainer
from datasets import load_dataset

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="unsloth/Qwen2.5-3B-Instruct",
    max_seq_length=2048,
    load_in_4bit=True,
)

dataset = load_dataset("AmareshHebbar/radiology-coder-sft", split="train")

trainer = SFTTrainer(
    model=model,
    tokenizer=tokenizer,
    train_dataset=dataset,
    dataset_text_field="messages",
    max_seq_length=2048,
)
trainer.train()

Related datasets in this collection

Dataset	Task	Rows
icd10-coder-sft	ICD-10-CM coding	74.7k
symptom-diagnoser-sft	Symptom → diagnosis	119k
clinical-summarizer-sft	SOAP summarization	30k
discharge-qa-sft	Discharge summary QA	30k
pmjay-classifier-sft	PM-JAY packages	11.1k
radiology-coder-sft	Radiology coding	25k
medical-ner-sft	Clinical NER	16.7k
hindi-medical-sft	Hindi medical QA	19.7k

Citation

@misc{axiomapper2026,
  author    = {Hebbar, Amaresh},
  title     = {AxisMapper: Medical AI Fine-tuning Dataset Suite},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/collections/AmareshHebbar/axiomapper-medical-ai-suite}
}

AxisMapper is an open-source project. Star the repo, open issues, and contribute at GitHub.