Motion Correction Model for Heart Images Using Combined Loss Function
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical imaging techniques face challenges in correcting motion artifacts in heart images due to the continuous motion of the heart, leading to poor image quality and instability in coronary artery correction, especially when conventional artifact correction approaches rely on global loss functions and require segmentation of coronary arteries.
Innovation Solution
A system and method for motion correction using a combined loss function that includes local, dice-related, and global loss functions to train a motion correction model, allowing for efficient and accurate correction of coronary artery artifacts without segmentation, by generating a motion correction model based on training samples and evaluating correction effects using a weighted sum of loss functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional artifact correction approaches use global loss functions, then the correction process is simplified, but the correction accuracy and stability deteriorate
Solution Approach 1:
The patent divides the correction process into multiple independent loss function components (local loss function for coronary artery regions, dice-related loss function for spatial overlap, and global loss function for overall image quality). Each component addresses specific aspects of the correction problem independently, allowing for more precise optimization than a single global loss function while maintaining manageable complexity through modular design.
2Measurement precision
If conventional methods require segmentation of coronary arteries, then the correction can be targeted, but the device complexity and processing time increase
Solution Approach 1:
The patent introduces a mask as an intermediary element that identifies coronary artery regions without requiring full segmentation. The mask serves as a simplified representation that guides the local loss function to focus on relevant areas, achieving targeted correction while avoiding the complexity of detailed segmentation procedures.
3Productivity
If a single loss function is used for training, then the training process is faster, but the correction stability and accuracy deteriorate
Solution Approach 1:
The training process is segmented into multiple loss function evaluations (local, dice-related, and global components) that are combined through a weighted sum. This multi-component approach maintains comprehensive training coverage for stability and accuracy while allowing parallel computation of individual components to preserve training efficiency.
4Device complexity
If heart motion is not corrected, then the imaging process is simpler, but the image quality and diagnostic value deteriorate
Solution Approach 1:
The patent replaces mechanical motion correction approaches with an artificial intelligence-based computational method. The motion correction model uses deep learning to automatically correct motion artifacts in the images, substituting complex mechanical correction systems with a software-based solution that achieves better correction效果 while maintaining operational simplicity.
Data Source
AI summary
Systems and methods for motion correction for a medical image. The systems may obtain a plurality of training samples each of which includes a sample image of a heart and a gold standard image of the heart. The sample image may have a motion artifact and the gold standard image may be with substantial removal of the motion artifact. The systems may also determine a motion correction model by training, based on the plurality of training samples according to a combined loss function, a preliminary model. The combined loss function may include at least a local loss function.