Automated Breast Arterial Calcifications Segmentation and Quantification on Mammograms Using Deep Learning



Software for the quantification of breast arterial calcifications on routine mammograms to be used for risk stratification for cardiovascular outcomes.

Key Benefits

  • Optimized image segmentation accuracy with reduced software complexity.
  • Integrated calcium scoring.
  • Screens patients who would not normally be indicated for cardiac CT.Potential to synergize and complement HPV E6/7-specific TCR therapies currently undergoing clinical trials.
  • Reduces costs associates with cardiac CT.

Market Summary

One in five women in the United States die from cardiovascular disease (CVD.) Women's symptoms often differ from men's, especially for heart attacks, making it particularly challenging to diagnose. Breast arterial calcification (BAC) is a medial calcification easily detected on standard mammography. Over the past decade, several studies show that BAC correlates with the risk of cardiovascular disease in women. Since millions of women undergo mammography, a significant relationship between BAC and CVD would provide an opportunity to improve risk stratification without additional cost and radiation exposure.

Technical Summary

The invention consists of a software designed for automatic detection and quantification breast arterial calcification (BAC) in female patients at risk for cardiovascular events and to assist in follow the progression of vascular calcifications while avoiding additional radiation exposure or cost to the patient. The software called Simple Context U-Net (SCU-Net) processes large image size of mammograms by breaking the image into smaller high-resolution patches. The software performs image segmentation to distinguish BAC from the rest of the image using reduced training parameters making it ideal for real-time clinical implementation. Furthermore, the soft quantified BAC using 5 metrics that produce a “calcium score” for each image and the associated vessel.

Patent Information

Tech ID: 21005
Published: 7/6/2022