Machine Learning in Intravascular Ultrasound: Validating Automated Lesion Assessment for Complex Coronary Interventions
Catheterization and Cardiovascular Interventions · 2025-05 · 105 (6) · 1320-1328
DOI: 10.1002/ccd.31458 ↗Kurz gefasst
Validierungsstudie in der Zeitschrift Catheterization and Cardiovascular Interventions, online erschienen am 21.02.2025. Verfasst von Jason Galo und 12 weiteren Autor:innen. Beteiligte Einrichtung: Section of Interventional Cardiology, MedStar Washington Hospital Center, Washington, District of Columbia, USA. Hauptthemen laut MEDLINE: Ultrasonography, Interventional, Machine Learning, Coronary Artery Disease, Coronary Vessels. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 13 Literaturangaben.
Abstract (Originalsprache)
Background
Intravascular ultrasound (IVUS) is essential for assessing complex coronary lesions, but remains underutilized in part due to difficulties in image interpretation. The AVVIGO IVUS Automated Lesion Assessment (ALA) software, which uses machine learning (ML) for automatic segmentation, promises to simplify lesion assessment. This study evaluated the agreement in stent size selection between ALA, an independent core laboratory (CL), and an expert interventional cardiologist (IC) for complex lesions.
Aims
The primary endpoint was the agreement in stent size selection, within 0.25 mm, of AVVIGO ALA automatic segmentation of Class I lesions against the gold‐standard measurement by an independent CL analysis and against an expert IC, (H. H.). The secondary endpoint was to assess the relative differences between AVVIGO ALA and CL, AVVIGO ALA and IC, and CL and IC, in vessel and lumen areas.
Methods
Patients with complex coronary lesions, including left main bifurcation, long, and severely calcified lesions, were retrospectively analyzed using IVUS with ALA. Stent size selection and area measurements by ALA were compared against a CL and IC using established sizing methods.
Results
In 48 patients, ALA demonstrated high agreement with CL (92%–100%) and IC (91%–98.5%) in stent size selection across lesion subtypes using recommended sizing methods. Lumen‐based sizing achieved higher agreement than vessel‐based sizing, particularly in calcified lesions (100% vs. 87%). The variability in relative difference in measurements between ALA and CL was greater than IC and CL in distal vessel and lesion vessel areas. The relative difference in measurements between ALA and IC was greater in vessel‐based sizing compared to lumen‐based sizing in the distal reference marker.
Conclusion
AVVIGO ALA demonstrated high agreement in stent size selection compared to a CL and expert IC. ML's ability to automate IVUS analysis may improve operator efficiency, reduce radiation exposure, and enhance the adoption of intravascular imaging in routine practice. It remains to be seen if it will impact of adoption of IVUS to guide complex PCI.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Section of Interventional Cardiology, MedStar Washington Hospital Center, Washington, District of Columbia, USA.
- Department of Internal Medicine, MedStar Washington Hospital Center, Washington, District of Columbia, USA.
- Cardiovascular Research Foundation, New York, New York, USA.
ORCID-Profile: Jason Galo, Ron Waksman
Zitieren
APA
Galo, J., Chaturvedi, A., Al-Qaraghuli, A., Rubio, P. M., Zhang, C., Kahsay, Y., Verma, B. R., Chitturi, K. R., Mintz, G. S., Waksman, R., Case, B. C., Hashim, H., & Garcia-Garcia, H. M. (2025). Machine Learning in Intravascular Ultrasound: Validating Automated Lesion Assessment for Complex Coronary Interventions. Catheterization and Cardiovascular Interventions, 105(6), 1320-1328. https://doi.org/10.1002/ccd.31458
Vancouver
Galo J, Chaturvedi A, Al-Qaraghuli A, Rubio PM, Zhang C, Kahsay Y, et al. Machine Learning in Intravascular Ultrasound: Validating Automated Lesion Assessment for Complex Coronary Interventions. Catheterization and Cardiovascular Interventions. 2025;105(6):1320-1328. doi:10.1002/ccd.31458
BibTeX
@article{Galo2025,
author = {Galo, Jason and Chaturvedi, Abhishek and Al-Qaraghuli, Abdullah and Rubio, Pablo M and Zhang, Cheng and Kahsay, Yirga and Verma, Beni Rai and Chitturi, Kalyan R and Mintz, Gary S and Waksman, Ron and Case, Brian C and Hashim, Hayder and Garcia-Garcia, Hector M},
title = {Machine Learning in Intravascular Ultrasound: Validating Automated Lesion Assessment for Complex Coronary Interventions},
journal = {Catheterization and Cardiovascular Interventions},
year = {2025},
volume = {105},
number = {6},
pages = {1320--1328},
doi = {10.1002/ccd.31458}
}
EUVIMED