Artificial Intelligence Algorithm for Subclinical Breast Cancer Detection.

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology and Cancer Research

Background:

• Early breast cancer detection significantly reduces morbidity and mortality.
• Current screening methods aim for timely diagnosis, but advancements are needed for risk stratification.
• Artificial intelligence (AI) shows promise in medical image analysis for disease detection.

Purpose of the Study:

• To evaluate if a commercial AI algorithm can predict future breast cancer development.
• To assess the utility of AI-derived scores in identifying women at high risk for breast cancer.
• To explore the potential for AI in personalizing breast cancer screening strategies.

Main Methods:

• Retrospective cohort study involving 116,495 women aged 50-69 with no prior breast cancer history.
• Utilized screening mammography data from 2004-2018 and an AI algorithm (INSIGHT MMG) for cancer detection scoring.
• Analyzed AI scores to compare women who developed screening-detected cancer, interval cancer, or no cancer over three biennial screenings.

Main Results:

• Mean absolute AI scores were significantly higher in breasts that developed cancer up to 4-6 years later compared to those that did not.
• The difference in AI scores between breasts that developed cancer and those that did not increased substantially with each screening round.
• Receiver operating characteristic analysis showed strong performance (AUC up to 0.96) for AI score differences in predicting screening-detected cancer.

Conclusions:

• Commercial AI algorithms for breast cancer detection can potentially identify women at elevated risk of future cancer.
• AI-driven risk assessment may enable personalized screening approaches, leading to earlier breast cancer diagnosis.
• These findings support the integration of AI into screening programs for improved patient outcomes.