medgemma-brain-cancer

🧠 MedGemma-Brain-Cancer

medgemma-brain-cancer is a fine-tuned version of google/medgemma-4b-it, trained specifically for brain tumor diagnosis and classification from MRI scans. This model leverages vision-language learning for enhanced medical imaging interpretation.

🔬 Model Details

• Base Model: google/medgemma-4b-it

• Dataset: orvile/brain-cancer-mri-dataset

• Fine-tuning Approach: Supervised fine-tuning (SFT) using Transformers Reinforcement Learning (TRL)

• Task: Brain tumor classification from MRI images

• Pipeline Tag: image-text-to-text

• Accuracy Improvement:

  • Base model accuracy: 33%
  • Fine-tuned model accuracy: 89%

📊 Results & Notebook

Explore the training pipeline, evaluation results, and experiments in the notebook:

👉 Fine_tuning_MedGemma.ipynb

🚀 Inference Example

# pip install transformers accelerate
from transformers import AutoProcessor, AutoModelForImageTextToText
from PIL import Image
import requests
import torch

model_id = "kingabzpro/medgemma-brain-cancer"

model = AutoModelForImageTextToText.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id)

# Example Brain MRI image — attribution: Orvile, via Kaggle dataset
image_url = "https://storage.googleapis.com/kagglesdsdata/datasets/7006196/11239552/Brain_Cancer%20raw%20MRI%20data/Brain_Cancer/brain_menin/brain_menin_0002.jpg?X-Goog-Algorithm=GOOG4-RSA-SHA256&X-Goog-Credential=databundle-worker-v2%40kaggle-161607.iam.gserviceaccount.com%2F20250527%2Fauto%2Fstorage%2Fgoog4_request&X-Goog-Date=20250527T102729Z&X-Goog-Expires=345600&X-Goog-SignedHeaders=host&X-Goog-Signature=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"
image = Image.open(requests.get(image_url, headers={"User-Agent": "example"}, stream=True).raw)

messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "text": None, "image": image},
            {"type": "text", "text": "What is the most likely type of brain cancer shown in the MRI image?\nA: brain glioma\nB: brain menin\nC: brain tumor"}
        ]
    }
]

inputs = processor.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=True,
    return_dict=True, return_tensors="pt"
).to(model.device, dtype=torch.bfloat16)

input_len = inputs["input_ids"].shape[-1]

with torch.inference_mode():
    generation = model.generate(**inputs, max_new_tokens=20, do_sample=False)
    generation = generation[0][input_len:]

decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)

Expected Output:

B: brain menin

🧪 Intended Use

This model is intended for research and educational purposes related to medical imaging, specifically brain tumor classification. It is not a certified diagnostic tool and should not be used in clinical decision-making without further validation.

🏷️ Tags

• medical
• brain_tumor
• mri
• trl
• sft

📜 License

Apache 2.0 License