This research adapts Contrastive Language-Image Pre-training (CLIP) models specifically for mammography, addressing the unique challenges of medical imaging with limited data resources.

• Introduces a specialized CLIP adaptation for mammography screening
• Implements multi-view alignment to leverage different breast perspectives
• Utilizes multi-scale techniques to capture details at various resolutions
• Demonstrates potential for improving breast cancer detection with fewer labeled examples

This work represents a significant step forward for medical AI, potentially enhancing breast cancer screening accuracy while reducing the need for extensive labeled datasets that have previously limited CLIP's application in specialized medical domains.

Multi-View and Multi-Scale Alignment for Contrastive Language-Image Pre-training in Mammography