Angiographic Frame Selection for Accurate Vascular Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vascular segmentation and feature identification in angiographic images are prone to errors due to low contrast and complex environments, leading to inaccurate three-dimensional models of the heart.
Innovation Solution
A system utilizing machine learning models, such as convolutional neural networks, to select an optimal angiographic image from a sequence based on contrast and quality scores, reducing the need for manual adjustments and enhancing the accuracy of downstream processes like three-dimensional model generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If automated vascular segmentation is performed on angiographic images, then the time and effort required for manual identification is reduced, but the accuracy decreases due to low contrast and complex environment
Solution Approach 1:
The patent applies segmentation by dividing the vascular tree identification task into multiple hierarchical levels: initial automated detection of vascular structures, followed by selective manual refinement of specific segments. The system segments the image processing workflow into automated low-level feature extraction and human-guided high-level verification, allowing most routine identification to be automated while maintaining accuracy for critical segments through targeted manual intervention.
Solution Approach 2:
The patent introduces an intermediary verification layer between automated segmentation and final vascular tree generation. This intermediary system provides confidence scoring for automated detections and routes uncertain cases for manual review, acting as a mediator that combines the speed of automated methods with the accuracy of manual verification only where needed.
2Measurement precision
If image resolution is increased to improve vascular feature visibility, then the accuracy of segmentation improves, but the radiant exposure and energy consumption increase
Solution Approach 1:
The patent applies partial action by processing only the most promising image frames for vascular segmentation rather than all available frames. The system selectively identifies and processes a subset of images that have optimal contrast and vascular visibility, avoiding the energy cost of processing every frame at high resolution while still capturing sufficient vascular information for accurate segmentation.
Solution Approach 2:
The patent dynamically adjusts processing parameters based on image quality assessment. Instead of uniformly applying high-resolution processing to all images, the system modifies processing intensity and resolution levels according to the specific characteristics of each frame, applying higher processing only to frames that meet quality thresholds and would benefit from enhanced analysis.
3Measurement precision
If manual verification of automated vascular segmentation is performed, then the accuracy improves, but the time and skill required increases
Solution Approach 1:
The patent applies local quality by directing manual verification efforts to specific local regions or segments of the vascular tree rather than requiring comprehensive manual review of the entire vascular structure. The system identifies areas with low confidence scores or anatomical complexity and targets manual verification specifically to those regions, allowing operators to focus their expertise where it is most needed while leaving high-confidence automated segments unchanged.
Data Source
AI summary
Methods for automated identification of vascular features are described. In some embodiments, one or more machine learning (ML)-based vascular classifiers are used, with their results being combined to with results of at least one other vascular classifier in order to produce the final results. Potentially advantages of this approach include the ability to combine certain strengths of ML classifiers with segmentation approaches based on more classical (“formula-based”) methods. These strengths may include particularly the identification of anatomically identified targets mixed within an image also showing similar looking but anatomically distinct targets.