HOPPR Launches Foundation Model to Speed Up AI Development for Mammography
HOPPR, a secure AI development platform for medical imaging, has announced the release of its latest product—the HOPPR™ EB 2D Mammography Foundation Model. This newly launched tool is designed to help developers build artificial intelligence applications for breast imaging more efficiently and with greater flexibility.
The model demonstrates strong performance in both binary and multi-class classification tasks, including cancer detection, breast density assessment, and device identification. Its adaptable architecture also enables it to be customized for a range of other downstream applications in mammography and related imaging workflows. HOPPR positions the model as an ideal foundation for research and operational solutions across the breast imaging field.
"We've been exploring ways to accelerate readiness of AI in breast imaging," said Sham Sokka, Chief Operating and Technology Officer at DeepHealth. "HOPPR's mammography foundation model should give us a flexible infrastructure to adapt it to our workflow needs. It's a meaningful step forward in accelerating development, readiness, and real-world use."
Timed with Breast Cancer Awareness Month, the announcement highlights the growing need for accessible AI infrastructure that fosters innovation in early detection and breast health.
The HOPPR™ EB 2D Mammography Foundation Model was built using a multi-stage training process involving self-supervised learning and expert-driven distillation. When developers fine-tune the model for specific classification tasks, it is provided with lightweight MLP heads and parameter-efficient LoRA adapters. Results are returned in standardized JSON format, simplifying integration into existing pipelines.
Key performance metrics from internal testing are impressive, with the model achieving an ROC-AUC of 0.92 for cancer detection, 0.94 for breast density classification, and 0.99 for pacemaker identification. The platform also enables fine-tuning for multiple classification levels—study, breast, or image—making it adaptable for a wide range of use cases.
HOPPR’s platform emphasizes developer control and transparency. Rather than offering prebuilt applications, it provides secure, HIPAA-compliant APIs for fine-tuning and inference, allowing users to train the model using their own labeled DICOM data. The system is built under a quality management framework that aligns with ISO 13485, SOC 2, and HITRUST standards, ensuring a secure environment for data handling and model development.
"Foundation models are changing the pace of innovation in imaging AI, but only if they're accessible, adaptable, and built with real-world deployment in mind," said Dr. Khan Siddiqui, CEO and Co-founder of HOPPR. "With this release, we're giving developers the infrastructure to move quickly with transparency, traceability, and control from day one."
The mammography model joins HOPPR’s growing suite of imaging AI tools, which already includes the HOPPR™ MC Chest Radiography Foundation Model. The company continues to expand its offerings with usage-based billing, secure APIs, and scalable development tools tailored to the evolving needs of medical imaging teams.