Regression-Based Mitral Valve Segmentation from 2D+t MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cardiac magnetic resonance (CMR) imaging techniques face challenges in accurately performing 4D anatomical and functional analysis of the heart valves due to 2D slice-based acquisition limitations, which hinder the extraction of detailed mitral valve information.
Innovation Solution
A regression-based method for 4D mitral valve segmentation from 2D+t MRI slices, utilizing a mitral valve-specific acquisition protocol and learning a regression model from data across different imaging modalities (CT, MRI, US) to estimate a patient-specific 4D mitral valve model, incorporating shape descriptors and additive boosting regression for robust segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If 2D slice-based CMR acquisition is used, then non-ionizing radiation and multi-plan ability are maintained, but 4D anatomical and functional analysis accuracy deteriorates due to long throughput times and acquisition limitations
Solution Approach 1:
The patent applies dimensionality change by reconstructing 4D mitral valve models (3D space + time) from 2D+t CMR slices. The regression-based method maps 2D contour information across multiple time points to generate 3D anatomical structures and their temporal evolution, effectively adding spatial and temporal dimensions to the analysis without requiring 4D acquisition
Solution Approach 2:
The patent introduces landmarks and regression models as intermediaries between 2D CMR slices and 4D mitral valve analysis. Landmarks serve as correspondence points across slices, while regression models predict 3D surface positions from 2D contour data, enabling accurate 4D reconstruction without direct 4D imaging
2Object-affected harmful factors
If 2D slice-based CMR acquisition is used, then non-ionizing radiation advantage is maintained, but productivity deteriorates due to long throughput times
Solution Approach 1:
The patent applies preliminary action by pre-training regression models using additive boosting on training data from multiple imaging modalities (CT, MRI, US) before actual CMR analysis. This pre-learning phase enables rapid inference during clinical use, reducing throughput time while maintaining accuracy
3Ease of manufacture
If traditional landmark-based 3D segmentation methods are used, then some mitral valve information can be extracted, but segmentation accuracy deteriorates due to complexity of heart valves and 2D acquisition limitations
Solution Approach 1:
The patent applies parameter changes by transforming the segmentation approach from direct 3D landmark-based methods to a regression-based framework that learns optimal mappings from 2D contour parameters to 3D surface parameters. The additive boosting regression dynamically adjusts prediction parameters based on training data, improving accuracy for complex valve structures
Solution Approach 2:
The patent uses copying by creating virtual 3D mitral valve surface models from 2D CMR slice contours through regression-based prediction. Instead of directly segmenting 3D data, the method generates accurate 3D surface copies that preserve anatomical details while overcoming 2D acquisition limitations
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A system and method for regression-based segmentation of the mitral valve in 2D+t cardiac magnetic resonance (CMR) slices is disclosed. The 2D+t CMR slices are acquired according to a mitral valve-specific acquisition protocol introduced herein. A set of mitral valve landmarks is detected in each 2D CMR slice and mitral valve contours are estimated in each 2D CMR slice based on the detected landmarks. A full mitral valve model is reconstructed from the mitral valve contours estimated in the 2D CMR slices using a trained regression model. Each 2D CMR slice may be a cine image acquired over a full cardiac cycle. In this case, the segmentation method reconstructs a patient-specific 4D dynamic mitral valve model from the 2D+t CMR image data.