README.md

metadata

license: apache-2.0
base_model: convnext_tiny_in22k
tags:
  - medical
  - healthcare
  - image-classification
  - brain-tumor-detection
datasets:
  - medical-images
language:
  - en
library_name: transformers
pipeline_tag: image-classification

Brain Tumor Detection

Model Description

This model is a ConvNeXt Tiny architecture trained with FastAI for detecting brain tumors in MRI scans.
It can classify brain MRI images as either showing signs of a tumor or being normal (no tumor detected).
Note: This model uses FastAI format and requires specific loading procedures.

Intended Uses & Limitations

⚠️ Important: This model is for research and educational purposes only. It should NOT be used for actual medical diagnosis without proper clinical validation and oversight by qualified medical professionals.

Intended Uses

• Research and development in medical AI
• Educational purposes for learning about medical image classification
• Proof-of-concept applications with proper disclaimers
• Academic studies and benchmarking

Limitations

• Not clinically validated
• Should not replace professional medical diagnosis
• May have biases based on training data
• Performance may vary on different populations or imaging conditions

Model Details

• Model Type: Image Classification
• Architecture: convnext_tiny_in22k
• Classes: 2
• Input: RGB images (224x224 pixels)

Classes

• No Tumor
• Tumor Detected

Usage

from transformers import AutoModelForImageClassification, AutoImageProcessor
from PIL import Image
import torch

# Load model and processor
model = AutoModelForImageClassification.from_pretrained("your-username/brain-tumor-detection")
processor = AutoImageProcessor.from_pretrained("your-username/brain-tumor-detection")

# Load and process image
image = Image.open("path_to_image.jpg")
inputs = processor(image, return_tensors="pt")

# Make prediction
with torch.no_grad():
    outputs = model(**inputs)
    predicted_class_id = outputs.logits.argmax().item()
    predicted_class = model.config.id2label[predicted_class_id]

print(f"Predicted class: {predicted_class}")

Training Details

This model was fine-tuned from pre-trained vision transformers on medical image datasets. For detailed training information, please refer to the original model documentation.

Evaluation

The model has been tested on held-out validation sets with the reported accuracy metrics. However, clinical evaluation and validation are required before any medical application.

Ethical Considerations

• Medical AI models can have significant impact on human health
• Proper validation and regulatory approval required for clinical use
• Potential for bias in training data and model predictions
• Should be used responsibly with appropriate medical oversight

Contact

For questions about this model, please create an issue in the repository.

Citation

If you use this model in your research, please cite appropriately and acknowledge that it's for research purposes only.

FastAI Usage

This model uses FastAI format. To use it:

from fastai.vision.all import load_learner
import pathlib
import platform

# Fix for cross-platform compatibility
if platform.system() == 'Windows':
    pathlib.PosixPath = pathlib.WindowsPath

# Load the model
model = load_learner('model.pkl')

# Make prediction
prediction, pred_idx, probs = model.predict(image)
print(f"Prediction: {prediction}")

Requirements

• fastai<2.8.0
• torch<2.7
• timm
• pathlib (for cross-platform compatibility)