Bias in Artificial Intelligence: Impact on Breast Imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Artificial intelligence (AI) is increasingly utilized in breast imaging to enhance efficiency and accuracy.
• Growing demand for breast imaging services strains limited physician resources.
• Real-world clinical settings present dynamic environments with diverse patient populations.

Purpose of the Study:

• To review the concept, sources, and types of AI bias in breast imaging.
• To offer strategies for mitigating AI bias in clinical practice.
• To ensure equitable adoption of AI in breast imaging.

Main Methods:

• Literature review on AI bias in medical applications.
• Analysis of AI model generalizability and adoption challenges.
• Synthesis of strategies for bias mitigation.

Main Results:

• AI models trained on specific datasets may not generalize to diverse populations.
• Potential for AI bias to negatively impact clinical outcomes.
• Identification of key sources and types of AI bias.

Conclusions:

• Addressing AI bias is crucial for the successful and equitable implementation of AI in breast imaging.
• Proactive strategies are needed to mitigate bias and ensure AI benefits all patient groups.
• Further research on bias detection and mitigation in AI for breast imaging is warranted.