Background: Women with intermediate breast cancer risk are not clearly represented in current screening guidelines. Magnetic Resonance Imaging (MRI) is a sensitive modality, yet its use in this group remains controversial due to cost and unclear benefit. Artificial Intelligence (AI) offers the potential to identify subgroups within the intermediate-risk category that could benefit most from MRI screening.

Objective: To assess the performance of an AI-based risk stratification model in identifying intermediate-risk women who would benefit from MRI screening.

Methods: We retrospectively analyzed mammographic, clinical, and genetic data from 15,000 women aged 40–70. Intermediate-risk was defined by Tyrer-Cuzick model scores (15–20%). An AI model incorporating imaging and non-imaging data was developed to stratify this population. The outcomes of MRI screening in AI-identified high-priority intermediate-risk women were compared to those managed by mammography alone.

Results: AI identified 1,200 women (8%) in the intermediate-risk cohort with an elevated likelihood of breast cancer development. In this subgroup, MRI detected significantly more cancers (CDR: 12.3 per 1,000) than conventional mammography (CDR: 5.6 per 1,000, p<0.001), with a lower false-positive rate (2.1% vs. 4.7%). The AI model had an AUC of 0.84, indicating strong discriminative ability.

Conclusion: AI-based risk stratification offers a promising approach for selectively applying MRI in intermediate-risk women, potentially enhancing cancer detection while optimizing resource utilization.