CT Perfusion Segmentation Using Follow-Up Trained Core-Penumbra Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for segmenting core and penumbra in ischemic stroke lesions using computed tomography perfusion datasets lack reliable ground truth data and fail to account for the correlation between these regions, leading to inconsistent and unreliable segmentation.
Innovation Solution
A computer-implemented method using a trained function that incorporates both perfusion parameter maps and the raw 4D CTP dataset, employing separate subfunctions for penumbra and core segmentation, and utilizing follow-up examination data with recanalization success indicators to generate accurate segmentation information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If threshold segmentation is applied based on perfusion parameter maps, then segmentation can be performed, but the reliability and consistency of segmentation results deteriorate due to lack of standardized thresholds and unreliable ground truth data
Solution Approach 1:
The patent uses follow-up examination data (ground truth) to create training datasets that copy the actual anatomical structures and lesion boundaries. This allows the machine learning model to learn from real clinical outcomes rather than arbitrary thresholds, improving segmentation reliability while maintaining ease of use through automated processing.
Solution Approach 2:
The patent transforms the segmentation approach from using fixed threshold parameters to using dynamic parameters learned from training data. The model learns optimal segmentation parameters from follow-up examination results, adapting to different patient cases and improving reliability without requiring manual threshold selection.
2Productivity
If traditional segmentation methods are used, then processing time is reduced, but the precision and accuracy of core and penumbra localization deteriorate
Solution Approach 1:
The patent performs preliminary training of the machine learning model using follow-up examination data before actual segmentation. This preliminary action creates a robust model that can quickly and accurately segment lesions in clinical practice, achieving both high processing speed and precision through pre-computed learning.
Solution Approach 2:
The patent replaces traditional mechanical threshold-based segmentation with a machine learning-based system. This substitution enables the system to achieve higher measurement precision through learned patterns while maintaining fast processing speeds through automated model inference, eliminating the need for manual parameter adjustment.
3Ease of operation
If separate segmentation approaches are used for core and penumbra, then segmentation can be performed, but the consistency and reliability deteriorate due to failure to account for correlations between regions
Solution Approach 1:
The patent merges the segmentation of core and penumbra into a unified machine learning model that processes both regions simultaneously. The model learns the correlations and spatial relationships between core and penumbra from follow-up examination data, producing consistent and reliable segmentation results for both regions together rather than separately.
Data Source
AI summary
A computer-implemented segmentation method for segmenting a core and a penumbra in a four-dimensional computed tomography perfusion dataset of ischemic tissue in an image region of a patient, includes determining at least one parameter map for at least one perfusion parameter from the computed tomography perfusion dataset; and using the at least one parameter map and the computed tomograph perfusion dataset as input data to a trained function to determine output data, the output data including segmentation information of the penumbra and the core.