AI Stroke Detection via Collateral Circulation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing ischemic stroke, particularly in identifying the type of large vessel occlusion (embolism or ICAS), are inadequate, leading to inappropriate treatment directions due to the inability to differentiate between the two before surgery, which can result in deteriorated conditions or unnecessary follow-up treatments.
Innovation Solution
A medical image-based system that uses AI models combining RNN and CNN layers to analyze angiography images, perfusion, and diffusion images to determine if collateral circulation is present, and identifies whether the occlusion is embolism or ICAS, further classifying it as branching-site or truncal-type occlusion based on circulation type, providing tailored treatment directions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional diagnostic methods (CT, MRI) are used to diagnose ischemic stroke, then the diagnosis can be performed, but the type of large vessel occlusion (embolism or ICAS) cannot be differentiated before surgery
Solution Approach 1:
The system performs preliminary analysis of collateral circulation using AI models before surgery to predict occlusion type. By analyzing angiography images, perfusion images, and diffusion images in advance, the system provides treatment guidance before the actual surgical intervention, enabling better preparation and decision-making
Solution Approach 2:
The patent introduces collateral circulation as an intermediary indicator to infer occlusion type. Since direct observation of occlusion type is difficult, the system uses collateral circulation patterns (visible in angiography images) as a mediator to indirectly determine whether the occlusion is embolism or ICAS, thereby resolving the information loss
2Productivity
If thrombolytic treatment is performed without identifying occlusion type, then treatment can be administered quickly, but the patient's condition may deteriorate due to inappropriate treatment
Solution Approach 1:
The system performs preliminary classification of occlusion type using AI analysis of medical images before treatment administration. This preliminary action enables healthcare providers to select appropriate treatment protocols (thrombolytic vs. other treatments) in advance, ensuring both quick administration and high reliability by matching treatment to the specific occlusion type
Solution Approach 2:
The system provides feedback to healthcare providers regarding the predicted occlusion type and recommended treatment direction based on AI analysis of collateral circulation patterns. This feedback loop enables informed decision-making that balances treatment speed with treatment appropriateness
3Measurement precision
If comprehensive medical image analysis is performed to identify occlusion type, then treatment accuracy is improved, but the complexity of the diagnostic system increases
Solution Approach 1:
The patent merges multiple AI models (RNN and CNN) into a unified diagnostic system that processes different types of medical images (angiography, perfusion, diffusion) simultaneously. By combining these models and integrating their outputs, the system achieves high classification accuracy while managing complexity through a cohesive architecture rather than separate independent systems
Solution Approach 2:
The diagnostic system is designed with multi-functionality to handle various types of medical images and provide comprehensive analysis for different occlusion types. The AI models are trained to recognize multiple patterns and provide unified treatment guidance, reducing the need for separate specialized systems for different diagnostic tasks
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present disclosure relates to a method, an apparatus, and a system for detecting and classifying an ischemic stroke based on a medical image. A medical image based ischemic stroke detecting and type classifying apparatus according to an aspect of the present disclosure includes an acquiring unit which collects images related to a brain of at least one patient; a detecting unit which determines whether the at least one patient is a large vessel occlusion patient, based on the collected image; a determining unit which determines whether a type of the large vessel occlusion is embolism or intracranial atherosclerosis (ICAS), when the at least one patient is a large vessel occlusion patient; and a diagnosing unit which provides treatment direction information which is applied differently according to the determined type of the large vessel occlusion.