Automated Coronary Angiogram Analysis via Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing coronary stenosis severity and left ventricular ejection fraction in coronary angiography are prone to high inter-observer variability and require significant manual input, leading to inconsistent results and increased procedural risks due to radiation and dye exposure.
Innovation Solution
A machine learning-based system utilizing deep neural networks for automated coronary angiography interpretation, including classification of angiographic projection angles, identification of primary anatomic structures, localization of coronary artery segments, and estimation of stenosis severity, which reduces manual input and enhances reproducibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If ad-hoc visual assessment is used for coronary stenosis severity determination, then the procedure is simple and quick, but inter-observer variability is high and reproducibility is poor
Solution Approach 1:
The patent replaces the mechanical visual assessment method performed by cardiologists with an automated machine learning system. The system uses deep neural networks to automatically analyze coronary angiogram images, classify projection angles, identify anatomical structures, and quantify stenosis severity, thereby eliminating human subjectivity while maintaining assessment speed.
Solution Approach 2:
The machine learning system performs self-service by automatically processing and interpreting coronary angiogram images without requiring manual intervention for measurement or analysis. The system independently completes the entire assessment workflow from image input to stenosis quantification, ensuring consistent and reproducible results.
2Measurement precision
If quantitative coronary angiography (QCA) is used to quantify coronary stenosis severity, then measurement precision is improved, but operator input requirements increase and time consumption increases
Solution Approach 1:
The patent replaces manual QCA operations with automated machine learning processing. The system automatically performs frame selection, reference object identification, vessel wall tracing, and stenosis measurement without requiring operator intervention at any step, thereby achieving high measurement precision while eliminating time loss associated with manual input.
Solution Approach 2:
The machine learning system performs preliminary actions by pre-processing the angiogram images to automatically identify optimal frames, locate reference objects, and segment vessel structures before performing the actual stenosis measurement. This preparation is done automatically and rapidly, eliminating the need for manual preliminary steps.
3Measurement precision
If left ventriculography is performed to determine left ventricular ejection fraction, then diagnostic accuracy is improved, but radiation exposure and dye consumption increase leading to increased morbidity
Solution Approach 1:
The patent applies multi-functionality by using the existing coronary angiogram images for dual purposes: both for assessing coronary artery stenosis and for determining left ventricular ejection fraction. The machine learning system extracts LVEF information from the same angiogram data that is already being used for coronary analysis, eliminating the need for separate left ventriculography and thereby reducing radiation and dye exposure while maintaining diagnostic accuracy.
Data Source
AI summary
Disclosed herein are methods and systems for angiography interpretation. In one particular embodiment a method includes: classifying the primary anatomic structure of one or more angiogram images of a first patient; classifying the projection angle of the one or more angiogram images of the first patient; labeling stenoses within the one or more angiogram images of the first patient classified as including a left or right coronary artery; producing one or more angiogram images of a second patient with corresponding estimated stenoses of the second patient to produce training data; training a machine learning model with the training data; estimating the arterial stenoses severity of the first patient by running the machine learning model on the filtered and labeled one or more angiogram images of the first patient, wherein the machine learning model is only run on angiogram images previously labeled as including stenoses.