Subject-Specific 3D MRA Synthesis From Multi-Contrast MRI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing in-silico studies of cerebrovascular haemodynamics are limited by the lack of large-scale databases with quantitative characterizations of whole-brain vasculature, particularly MRA images, which are not included in standard scanning protocols due to long acquisition times, and public databases like BraVa contain insufficient samples for inter-patient variability analysis.
Innovation Solution
A method is developed to synthesize subject-specific 3D MRA images from multi-contrast MR images using a conditional Generative Adversarial Network (cGAN) that preserves vascular anatomy, enabling the creation of large-scale digital twin cohorts for in-silico studies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MRA images are acquired using standard scanning protocols, then quantitative characterization of whole-brain vasculature is achieved, but acquisition time becomes excessively long
Solution Approach 1:
The patent synthesizes MRA images by copying and transforming information from routinely acquired multi-contrast MR images (T1, T2, PD-weighted) to generate subject-specific MRA images. This creates a copy of the vascular information that would otherwise require separate lengthy MRA acquisition, thereby resolving the contradiction between obtaining quantitative vascular characterization and minimizing acquisition time
Solution Approach 2:
The patent performs preliminary synthesis of MRA images from routinely acquired multi-contrast images during standard scanning protocols. By preparing and synthesizing the MRA images in advance from already-acquired data, the system eliminates the need for separate time-consuming MRA acquisition while ensuring quantitative vascular characterization is available when needed
2Reliability
If large-scale databases with MRA images are created for in-silico studies, then inter-patient variability analysis is improved, but data collection and processing complexity increases
Solution Approach 1:
The patent creates a universal synthesis framework that works across multiple subjects and imaging scenarios. The cGAN model is trained on diverse multi-contrast MR images from different subjects and can generalize to synthesize MRA images for new subjects, enabling large-scale database creation without subject-specific customization. This universality reduces the complexity of data collection and processing while improving the reliability of inter-patient variability analysis
Solution Approach 2:
The patent enables efficient creation of large-scale databases by synthesizing MRA images from routinely acquired multi-contrast images for multiple subjects. This copying approach avoids the need to collect and store separate MRA acquisitions for each subject, significantly reducing data collection complexity while maintaining the ability to perform robust inter-patient variability analysis across large cohorts
Data Source
AI summary
There is provided a computer-implemented method for synthesising magnetic resonance angiography (MRA) images from other types of inputted magnetic resonance (MR) images, in a subject-specific manner, the method comprising providing a conditional generative adversarial network (cGAN) that learns a combined latent representation of the inputted magnetic resonance images for each subject and learns to transform this combined latent representation to a magnetic resonance angiography image corresponding to that subject, providing a plurality of magnetic resonance (MR) images as input into the cGAN, and outputting a plurality of MRA images from the cGAN based on the plurality of inputted MR images.