dermaintel-wound-classifier

DermaIntel Wound Type Classifier

Fine-tuned Vision Transformer (ViT) for medical wound type classification. Part of the DermaIntel medical imaging platform.

Model Description

This model classifies wound images into 7 categories:

• Pressure Ulcer/Injury - Caused by prolonged pressure on skin
• Venous Ulcer - Caused by venous insufficiency
• Arterial Ulcer - Caused by arterial insufficiency
• Diabetic Ulcer - Foot ulcers in diabetic patients
• Surgical Wound - Post-operative wounds
• Traumatic Wound - Injury-related wounds
• Normal/Healthy Skin - No wound detected

Intended Use

• Clinical decision support for wound assessment
• Automated wound documentation
• Wound care management in healthcare settings
• Educational and research purposes

Training

• Base Model: google/vit-base-patch16-224-in21k
• Fine-tuning: Custom wound image dataset
• Architecture: Vision Transformer (ViT-Base)
• Input Size: 224x224 RGB images

Usage

from transformers import ViTForImageClassification, ViTImageProcessor
from PIL import Image

# Load model
processor = ViTImageProcessor.from_pretrained("PayamFard123/dermaintel-wound-classifier")
model = ViTForImageClassification.from_pretrained("PayamFard123/dermaintel-wound-classifier")

# Classify image
image = Image.open("wound_image.jpg")
inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)
predicted_class = outputs.logits.argmax(-1).item()

Classes

Limitations

• This model is intended for clinical decision support, not replacement of clinical judgment
• Performance may vary with image quality and lighting conditions
• Should be used in conjunction with clinical assessment
• Not FDA approved for diagnostic use

Citation

If you use this model, please cite:

@misc{dermaintel-wound-classifier,
  title={DermaIntel Wound Type Classifier},
  author={DermaIntel Team},
  year={2024},
  publisher={HuggingFace},
  url={https://huggingface.co/PayamFard123/dermaintel-wound-classifier}
}

Contact

For questions or issues, contact the DermaIntel team.