New research published in Nature Medicine highlights the presence of bias in artificial intelligence (AI) machine learning models used in medical imaging. While these models excel at predicting race and gender, they do not perform as effectively in diagnosing diseases. The study emphasizes the importance of regular evaluation and offers best practices to address the disparity.

The use of AI in medical imaging is growing rapidly, with the market expected to reach USD 8. 18 billion by 2030. However, biases can impact accuracy, and factors such as training data quality and human biases can contribute to this issue. The researchers recommend careful assessment of AI models on in-distribution data to mitigate biases across different demographics.