Summary
A fully automatic, machine learning (ML) algorithm that segments lumen boundaries enabling a reliable and easily reproducible translation for clinical usage from IVUS images in patients with coronary artery disease (CAD).
Inventors
Hector Garcia-Garcia, MD, PhD
Gonzalo D. Ares
Pablo J. Blanco
Paul G.P. Ziemer
Gonzalo D. Maso Talou
What is it? What does it do?
Intravascular Ultrasound (IVUS) is the industry exemplar in imaging modalities for the assessment of coronary artery disease (CAD). One of the most arduous tasks when analyzing IVUS datasets is the delineation of the lumen, external elastic membrane (EEM) interface, and plaque. Traditionally, an expert manually outlines the structure or regions of interest (ROI). However due to the observed inter- and intra-variability in the tracings and the time-consuming nature of the process, this diagnostic tool is unreliable in a clinical setting.
A novel algorithm for automatic lumen segmentation focusing on the minimum lumen area (MLA) and percentage area of stenosis from IVUS datasets has been developed. These features are critical for clinical decision-making for the determination of the necessity of lesion treatment. The algorithm combines automatic gating, multi-frame (MF) convolutional neural network segmentation, and gaussian process regression which addresses longitudinal and transversal coherence encountered in IVUS datasets, making it a highly reliable and clinically useful tool.
Why is it better?
This novel algorithm:
-
Achieves fast, accurate, and precise assessment of the target lumen.
-
Allows for a reproducible computational technique with consistent results for clinical usage.
-
Eliminates the need for trained personnel.
What is its current status?
The novel, automatic segmentation algorithm was developed using gating, segmentation, and filtering techniques. The multi-frame (MF) convolutional neural network permitted not only a single IVUS frame but also its neighboring frames, improving the accuracy of segmentation performance. In-house software has been developed to implement the entire workflow process.
MedStar Inventor Services performed a patentability assessment and a PCT (US2022/035514) was filed based on their findings.
The MedStar Inventor Services team is now seeking a licensing/collaboration partner to help advance and commercialize this technology. Please contact us at invent@medstar.net.
Publications
-
Ziemer, P. G., Bulant, C. A., Orlando, J. I., Maso Talou, G. D., Alvarez, L. A. M., Guedes Bezerra, C., ... & Blanco, P. J. (2020). Automated lumen segmentation using multi-frame convolutional neural networks in intravascular ultrasound datasets. European Heart Journal-Digital Health, 1(1), 75-82.