Applications of artificial intelligence-based computer-assisted diagnosis in breast radiology: a narrative review

Abstract

Purpose: Mammography is the standard screening method for breast cancer, proven to reduce mortality. However, its diagnostic performance varies depending on patient characteristics and radiologist expertise. Dense breast tissue, present in approximately 70% of Korean women aged 40 to 59, limits detection by obscuring malignancies. Additionally, optimal interpretation requires extensive training, which is not always achievable. Artificial intelligence-based computer-aided diagnosis (AI-CAD) has emerged as a promising tool for enhancing mammographic accuracy and efficiency. Current Concepts: AI-CAD has shown diagnostic performance comparable to that of experienced radiologists while addressing the limitations of traditional CAD systems, particularly excessive false positives. Studies suggest AI-CAD improves radiologists' accuracy, particularly among those with limited breast imaging experience. In Europe, AI-assisted reading is increasingly recognized as a viable alternative to traditional double reading. In Korea, adoption of AI-CAD is expanding, with systems approved by the Korean Food and Drug Administration currently in clinical use. Recently, one AI-CAD system received conditional non-reimbursement designation, allowing hospitals to use it for up to 5 years while collecting clinical evidence to support future insurance coverage decisions. Discussion and Conclusion: AI-CAD has significant potential to enhance early breast cancer detection while maintaining acceptable false-positive rates, making it a valuable adjunct in screening programs. Beyond improved detection, AI-CAD may optimize workflow efficiency by triaging cases and prioritizing high-risk examinations. However, its integration into clinical practice necessitates standardized guidelines, regulatory oversight, and further validation through large-scale prospective studies. As AI technology continues to advance, ongoing investigation into its role in personalized breast cancer screening is essential.