MRI Contrast Synthesis for Faster Multi-Contrast Scanning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The prolonged scanning time for multi-contrast MRI exams can result in image artifacts or misalignment issues, and the use of reduced contrast agent doses can lead to poor-quality images, making it desirable to synthesize missing or low-quality contrast-weighted images.
Innovation Solution
A deep learning-based framework integrates anatomy and pathology information to synthesize contrast-weighted images using a segmentation network, classification network, and reconstruction network, allowing for faster MR acquisitions by generating images with different brightness levels and improving image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If multi-contrast MRI imaging is performed with all contrast types, then comprehensive anatomical and pathological information is obtained, but scanning time is prolonged (20-40 minutes)
Solution Approach 1:
The patent uses deep learning models to generate synthetic copies of contrast-weighted images that are not directly acquired. The model learns the mapping between different contrast types from training data and generates T1, T2, or other contrast images from acquired sequences, providing comprehensive information without acquiring all contrasts directly, thus reducing scan time while maintaining diagnostic quality
Solution Approach 2:
The deep learning model is pre-trained on large datasets of multi-contrast MRI images to learn the complex relationships between different contrast types. This preliminary training enables the model to accurately synthesize missing contrasts during actual clinical scanning, allowing rapid generation of comprehensive multi-contrast information from limited acquired sequences
2Quantity of substance
If contrast agent dose is reduced, then patient safety and cost are improved, but image quality deteriorates
Solution Approach 1:
The deep learning model acts as an intermediary that enhances the utility of low-dose contrast imaging. By learning the characteristic patterns of high-quality contrast images from training data, the model can process and enhance low-dose images or synthesize missing contrasts, effectively bridging the gap between reduced contrast agent usage and maintained image quality
Solution Approach 2:
The patent changes the approach from directly acquiring high-quality images with high contrast dose to using computational processing to achieve high-quality results. The model adjusts image parameters through learned transformations, enabling reconstruction of high-quality contrast images from low-dose inputs by leveraging statistical patterns learned from abundant training data
3Productivity
If scanning time is reduced by acquiring only selected contrasts, then productivity is improved, but completeness of diagnostic information is reduced
Solution Approach 1:
The system generates synthetic copies of unacquired contrast images through deep learning. When only T1 and T2 images are acquired, the model synthesizes other required contrasts (such as FLAIR, DWI, or other sequence types) by learning from training data, ensuring complete diagnostic information is available without extending scan time
Solution Approach 2:
The deep learning model serves multiple functions: it can synthesize various types of contrast images, perform image enhancement, correct artifacts, and adapt to different anatomical regions and pathological conditions. This multi-functionality allows a single acquisition protocol to yield comprehensive multi-contrast information through computational generation
Data Source
AI summary
Methods and systems are provided for synthesizing a contrast-weighted image in Magnetic resonance imaging (MRI). The method comprises: receiving a multi-contrast image of a subject, where the multi-contrast image comprises one or more images of one or more different contrasts; and generating, by a deep learning model, a synthesized image having a target contrast that is different from the one or more different contrasts of the one or more images. The deep learning model is trained by a framework comprising a segmentation network for generating a segmentation map, a classification network for generating a pathology aware map and a reconstruction network for generating a plurality of synthesized images with different brightness levels in a tissue area.