Machine Learning Vessel Abnormality Detection in CT Angiography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current imaging techniques for diagnosing abnormalities in coronary arteries, such as echocardiography and catheter angiography, are limited by their invasive nature, operator dependency, and inability to accurately detect deep-seated or progressing abnormalities like aneurysms in children with Kawasaki disease.
Innovation Solution
A computer-implemented method using machine learning algorithms to process computed tomography angiographic images, enabling the automatic detection and classification of abnormalities in coronary artery segments, thereby overcoming the limitations of existing techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If echocardiography is used to diagnose coronary artery abnormalities, then it is non-invasive and can be repeated frequently, but it is highly operator dependent and has poor sensitivity in demonstrating intra-luminal thrombus or arterial stenosis
Solution Approach 1:
A machine learning-based automated detection system serves as an intermediary between the echocardiography imaging process and diagnostic interpretation. The system processes echocardiographic images to automatically detect coronary artery abnormalities, thrombus, and stenosis, reducing operator dependency while maintaining the non-invasive nature of echocardiography.
Solution Approach 2:
The patent replaces the manual mechanical interpretation process by radiologists with an automated machine learning system. This substitution enhances detection sensitivity and consistency without requiring invasive procedures, as the algorithm objectively analyzes echocardiographic image data to identify abnormalities.
2Measurement precision
If catheter angiography is used to evaluate coronary artery abnormalities, then it provides detailed vascular imaging, but it is invasive, requires sedation, and is associated with high radiation exposure
Solution Approach 1:
The patent creates a virtual copy of the catheter angiography imaging capability using machine learning algorithms that process non-invasive echocardiographic images. This virtual reconstruction provides detailed vascular imaging information without requiring actual catheter insertion or high-radiation exposure, effectively copying the diagnostic value of angiography through software-based image analysis.
Solution Approach 2:
The patent substitutes the physical catheter angiography procedure with an automated image analysis system that processes echocardiographic data. This replacement eliminates the need for invasive catheter insertion, sedation, and high radiation exposure while maintaining the ability to detect detailed vascular abnormalities through computational analysis.
3Object-affected harmful factors
If MRA is used to image coronary arteries, then it has zero radiation exposure, but temporal and spatial resolution is limited and acquisition time is long
Solution Approach 1:
The patent applies preliminary machine learning-based image processing and enhancement techniques to echocardiographic images to compensate for their inherent limitations. By pre-processing the images with automated detection algorithms, the system extracts high-resolution vascular information that would otherwise be unavailable, achieving diagnostic quality without the long acquisition times of MRA.
Solution Approach 2:
The patent replaces the MRA imaging modality with an automated analysis system that processes echocardiographic images. This substitution maintains the zero radiation exposure advantage while overcoming the resolution limitations through computational enhancement, effectively replacing the need for time-consuming MRA scans with rapid automated image processing.
4Measurement precision
If manual review of CT angiographic images is performed, then abnormalities can be detected, but it is time-consuming and subject to human error
Solution Approach 1:
The patent implements a self-service automated detection system that independently analyzes CT angiographic images without requiring manual review by radiologists. The machine learning algorithm autonomously identifies abnormalities, measures vessel dimensions, and generates diagnostic reports, freeing up medical professionals from time-consuming manual image analysis while maintaining or improving detection accuracy.
Solution Approach 2:
The patent substitutes the manual mechanical review process with an automated machine learning system that rapidly processes CT angiographic images. This replacement eliminates human error and subjectivity while reducing diagnostic time from minutes or hours to seconds, as the algorithm simultaneously analyzes multiple image parameters and generates comprehensive assessments without fatigue or distraction.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Techniques for processing one or more computed tomography angiographic images are disclosed. The processing may take place either during or after a computed tomography angiography exam of an anatomical region of interest. The one or more computed tomography angiographic images are processed using at least one machine learning algorithm, and a presence of at least one abnormality associated with one or more vessel segments comprised in the anatomical region of interest is determined, detected, or diagnosed.