Transparent Arterial Layer Separation in Cardiac Angiography with Self-Supervised Deep Learning

NU 2023-092

INVENTORS
Sameer Ansari*
Donald Cantrell*
Leon Cho
Syed Hasib Akhter Faruqui
Chaochao Zhou

SHORT DESCRIPTION
This technology provides cardiac angiography with markedly enhanced visualization of the vasculature by leveraging self-supervised deep neural network-based background subtraction algorithms.

BACKGROUND
Cardiac angiography relies on catheter-based imaging to diagnose vascular disorders with high precision. Current standard-of-care cardiac angiographic images are degraded by superimposed densities from bones and soft tissues, along with cardiac motion.  These imaging limitations reduce vascular clarity and may necessitate repeat acquisitions that increase patient radiation exposure. Despite efforts to enhance vascular clarity and improve background subtraction and registration through various algorithms, existing approaches struggle to balance anatomic vascular fidelity and artifact suppression, underscoring the need for more robust and efficient solutions.

ABSTRACT
This technology introduces a deep neural network-based digital subtraction angiography (DSA) algorithm designed to improve the accuracy of vessel structure recovery and intensity information from cardiac angiography acquisitions. Utilizing all frames of cardiac angiograms, the model identifies inter-frame spatio-temporal connectivity amidst complex and noisy backgrounds, effectively separating and extracting vessels from obstructions such as the beating heart, bones, and tissue artifacts. This translational technology can be similarly applied to cardiac angiography.

APPLICATIONS

• Clinical Diagnostic Imaging Software Integration
  • Provides superior imaging for coronary angiography.

ADVANTAGES

• State-of-the-Art Vascular Enhancement with Advanced Deep Learning
  • Enables clearer separation of vessel structures.
• Cost and Radiation Exposure Reduction
  • Higher quality scans reduces frequency and cost of repeat scans and limits radiation exposure.
• Translational Technology
  • Adaptable algorithms can be applied to cerebral and peripheral vasculature imaging.

PUBLICATIONS
Cantrell, D.R., Cho, L., Zhou, C. et al. Background Subtraction Angiography with Deep Learning Using Multi-frame Spatiotemporal Angiographic Input. J Digit Imaging. Inform. Med. 37, 134–144 (2024). 

Zhou, C., Abdalla, R.N., et al. Reducing Motion Artifacts in Craniocervical Background Subtraction Angiography with Deformable Registration and Unsupervised Deep Learning. Radiology Advances. 1:3, umae020 (2024).

IP STATUS

• PCT Application
• US Application