International study highlights demographic biases in AI models diagnosing skin diseases from images.
Key Details
- 1Researchers evaluated ChatGPT-4 and LLaVA on 10,000 dermatoscopic images of skin diseases.
- 2Study assessed diagnostic accuracy and fairness regarding sex and age groups.
- 3ChatGPT-4 showed better demographic fairness than LLaVA, which had marked sex-based biases.
- 4Both AI models outperformed traditional deep learning approaches overall.
- 5Calls made for considering demographic fairness before clinical deployment of AI in healthcare.
- 6Further research planned to evaluate impact of skin tone and other demographic factors.
Why It Matters
Addressing bias in AI diagnostic tools is essential to ensure equitable healthcare outcomes. This study provides critical insights and grounds for improvement in AI model development for medical imaging.
Source
EurekAlert
Related News
•EurekAlert
AI Models Use EMR and Radiology Data To Predict Intimate Partner Violence
Mass General Brigham researchers developed AI models that use EMR, including radiology data, to predict risk for intimate partner violence (IPV) years before patients seek care.
•EurekAlert
AI Fusion Model Uses Radiology Data to Predict Intimate Partner Violence Risk
Researchers created an AI tool leveraging clinical and radiology data to accurately predict patients at risk of intimate partner violence (IPV).
•EurekAlert
MIT Researchers Advance Explainable AI for Medical Imaging
MIT and collaborators developed a technique to make computer vision models, including those used in medical imaging, provide clearer, concept-based explanations for their predictions.