Probabilistic Vessel Tree Tracing for Stroke Occlusion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional AI-based approaches for detecting large vessel occlusions in medical imaging are not robust in the presence of signal dropout, noise, vessel tortuosity, calcification, and proximity to bone or bifurcations, and fail to accurately identify the exact location of occlusions within vessel distribution models.
Innovation Solution
A probabilistic tree tracing method using reinforcement learning agents to navigate anatomical landmarks, generating a probabilistic tree of vessels based on geometric and local image context features, and detecting large vessel occlusions by identifying discontinuities in the tree.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional AI-based approaches are used for vessel segmentation, then the process is automated and decision-making is accelerated, but the robustness is reduced in the presence of signal dropout, noise, vessel tortuosity, calcification, and proximity to bone or bifurcations
Solution Approach 1:
The patent introduces a probabilistic tree model as an intermediary framework that integrates multiple AI-based segmentation results. This model represents vessel structures with probability distributions rather than deterministic boundaries, allowing it to handle uncertainties from signal dropout, noise, and anatomical variations. The probabilistic representation acts as a mediator that reconciles conflicting segmentation outcomes from different AI approaches while maintaining robustness against imaging artifacts and anatomical complexities.
Solution Approach 2:
The patent combines multiple AI-based segmentation approaches into a composite probabilistic model. By integrating results from different AI algorithms and representing them as probability distributions, the system creates a more robust segmentation that leverages the strengths of individual approaches while compensating for their weaknesses. This composite probabilistic representation improves reliability in the presence of signal dropout, noise, and vessel tortuosity.
2Extent of automation
If conventional AI-based approaches are used for LVO detection, then automatic interpretation is achieved, but the exact location of occlusion within vessel distribution models cannot be identified
Solution Approach 1:
The patent segments the vessel distribution model into discrete probabilistic tree structures, where each branch and node represents a specific vessel segment with associated probability distributions. This segmentation allows the system to automatically identify and localize occlusions within specific vessel segments rather than providing only a general LVO detection. The tree structure enables precise mapping of occlusion locations to anatomical vessel segments while maintaining automatic interpretation.
Solution Approach 2:
The patent replaces traditional mechanical or rule-based occlusion localization methods with a probabilistic AI-based approach. By using probabilistic tree tracing and reinforcement learning agents, the system automatically infers occlusion locations based on probability distributions derived from multiple AI segmentation results. This substitution maintains high automation while achieving precise occlusion localization through statistical inference rather than deterministic rules.
3Reliability
If multiple AI-based approaches are integrated to improve robustness, then segmentation reliability is improved, but the system complexity increases
Solution Approach 1:
The patent transforms the complexity issue by changing the representation parameter from deterministic binary segmentation masks to probabilistic distributions. This parameter change allows the system to integrate multiple AI approaches by combining their probability outputs rather than managing multiple complex segmentation pipelines. The probabilistic framework provides a unified mathematical language that simplifies the integration process while improving robustness through ensemble methods.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and methods for generating a probabilistic tree of vessels are provided. An input medical image of vessels of a patient is received. Anatomical landmarks are identified in the input medical image. A centerline of the vessels in the input medical image is determined based on the anatomical landmarks. A probabilistic tree of the vessels is generated based on a probability of fit of the anatomical landmarks and the centerline of the vessels. The probabilistic tree of the vessels is output.