IVUS-OCT Vessel Imaging With Reference-Frame Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting features in blood vessel images using intravascular ultrasound (IVUS) and optical coherence tomography (OCT) lack the ability to efficiently identify specific objects like stents and plaques with precision, as they uniformly process all images without considering variations.
Innovation Solution
A medical system utilizing a dual-type catheter with IVUS and OCT capabilities, combined with a machine learning model, selectively processes cross-sectional images to identify and highlight specific objects such as stents and plaques by determining a reference image and analyzing adjacent images for enhanced detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all cross-sectional images are uniformly processed by the detection method, then comprehensive coverage of the blood vessel is achieved, but detection efficiency and precision are reduced due to lack of selective analysis
Solution Approach 1:
The patent applies local quality by differentiating the processing approach for different regions of the blood vessel. Reference images are selected based on specific criteria (objects of interest, abnormal regions, or representative characteristics) and receive focused detailed analysis, while other images are processed more efficiently. This allows high detection precision for critical regions without uniformly processing all images, thereby improving overall detection efficiency.
Solution Approach 2:
The patent segments the blood vessel images into different categories: reference images requiring detailed analysis and non-reference images for efficient processing. By dividing the image set and applying different processing strategies to each segment, the system achieves both high detection precision for important features and improved overall detection efficiency.
2Measurement precision
If machine learning models analyze every generated blood vessel image, then detection accuracy is improved, but processing time and computational resources increase significantly
Solution Approach 1:
The patent implements preliminary action by pre-selecting reference images from the generated blood vessel images based on specific criteria before applying the machine learning detection model. This preliminary selection step identifies images containing objects of interest, abnormal regions, or representative characteristics, thereby reducing the total number of images requiring computationally intensive machine learning analysis and significantly decreasing processing time while maintaining high detection accuracy for critical features.
Solution Approach 2:
The patent changes the parameter of image selection by introducing a reference image selection mechanism that filters the input image set. By modifying which images are subjected to machine learning analysis (changing from all images to selectively chosen reference images), the system reduces computational load and processing time while preserving detection accuracy for the most important diagnostic features.
3Reliability
If uniform feature detection is applied to all images, then consistency across the entire blood vessel is maintained, but specific objects like stents and plaques may be missed due to lack of focused analysis
Solution Approach 1:
The patent applies local quality by directing focused detailed analysis specifically at reference images that contain objects of interest, abnormal regions, or representative characteristics. This localized concentrated analysis improves object identification precision for stents, plaques, and other critical features without compromising the overall reliability of the detection system, as the uniform processing of non-reference images maintains consistency across the entire blood vessel.
Data Source
AI summary
A medical system includes a catheter that includes a sensor and insertable into a luminal organ and an image processing apparatus configured to: generate first cross-sectional images of the organ based on sensor signals output when the catheter is moved along the organ, select second images from the first images at predetermined intervals, input the second images to a machine learning model and acquire a type and region of an object in each second image, determine one second image as a reference, determine two or more first images generated before and after the reference, input said two or more first images to the model and acquire the type and region of the object in said two or more first images, and output information based on the type and region of the object acquired from each of the reference and said two or more first images.