Convolutional Neural Network for Medical Motion Artifact Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical imaging technologies face challenges in effectively compensating for motion artifacts, which negatively affect the standard reconstruction and segmentation of images, particularly in areas like the coronary arteries, due to the integration of measurement signals over time, leading to difficulties in imaging resolution and increased radiation or magnetic field loads.
Innovation Solution
A method utilizing a convolutional neural network trained with a dataset of medical images to automatically identify and compensate for motion artifacts, allowing for the recognition and reduction of these artifacts in recorded medical images, thereby improving image quality without the need for shortening recording time or increasing radiation or magnetic field exposure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the recording time is shortened to reduce motion artifacts, then the motion artifacts are reduced, but the signal-to-noise ratio deteriorates and the radiation load increases
Solution Approach 1:
The patent applies motion artifact information, which was previously harmful to image quality, as training data for the neural network. The system learns to recognize and correct motion artifacts by training on images containing these artifacts, converting the harmful effect into a beneficial learning opportunity that improves subsequent image reconstruction accuracy
Solution Approach 2:
The patent performs preliminary reconstruction of images containing motion artifacts before final reconstruction. By first creating a motion artifact-containing image and using it as input for the neural network, the system prepares the data in advance for correction, enabling the final reconstruction to achieve higher quality without requiring shorter recording times
2Reliability
If the recording time is shortened to reduce motion artifacts, then the motion artifacts are reduced, but the radiation load increases
Solution Approach 1:
The patent converts the harmful effect of motion artifacts into beneficial training data. By training the neural network on images containing motion artifacts, the system learns to recognize and correct these artifacts, enabling high-quality imaging without the need to shorten recording time and thereby avoiding increased radiation exposure
3Productivity
If standard reconstruction and segmentation are performed, then the imaging process is completed, but motion artifacts negatively affect the results particularly in coronary arteries
Solution Approach 1:
The patent introduces a neural network as an intermediary between standard reconstruction and final segmentation. This neural network processes the reconstructed images to remove motion artifacts before segmentation is performed, thereby improving segmentation accuracy without compromising imaging process efficiency
Solution Approach 2:
The patent performs preliminary motion artifact correction using the neural network before conducting segmentation. By removing motion artifacts in advance, the subsequent segmentation process operates on cleaner data, improving accuracy particularly in challenging areas like coronary arteries
Data Source
AI summary
A method is for training a convolutional neural network of a compensation unit. The method includes: provisioning a machine learning device, the machine learning device being designed for training the convolutional neural network; provisioning a start compensation unit, including an untrained convolutional neural network, on or at the machine learning device; provisioning a training image dataset including a plurality of medical training input images and at least one training output image, wherein a reference object is shown essentially without motion artifacts in the at least one training output image and the reference object concerned is contained in the plurality of medical training input images with different motion artifacts; and training the convolutional neural network of the compensation unit in accordance with a principle of machine learning, using the training image dataset. A compensation unit, a machine learning device, a control device for controlling a medical imaging system are also disclosed.