Diffusion-weighted magnetic resonance imaging
Inventors
Blackledge, Matthew • Zormpas-Petridis, Konstantinos
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12189010-B2
Publication Date
2025-01-07
Expiration Date
Abstract
A method of performing diffusion-weighted magnetic resonance imaging is provided. The method includes a step of: using a neural network to filter a diffusion-weighted image of an object acquired by a magnetic resonance imaging scanner, the neural network being programmed to produce an output image from the acquired image. The neural network improves the signal to noise ratio of the output image relative to the acquired image. The neural network, when applied to a synthetic knife-edge image to which Rician noise providing a signal-to-noise ratio of 13 or more is added, forms a curve of normalised values of modulation-transfer-function against frequency which has a higher area thereunder than the area under the corresponding curve of normalised values of modulation-transfer-function against frequency for a reference Gaussian smoothing filter.
Core Innovation
The disclosed invention performs diffusion-weighted magnetic resonance imaging by using a neural network to filter a diffusion-weighted image of an object acquired by a magnetic resonance imaging scanner. The neural network is programmed to produce an output image from the acquired image, and the invention improves the signal to noise ratio of the output image relative to the acquired image.
The invention further specifies a performance requirement in which, when the neural network is applied to a synthetic knife-edge image to which Rician noise providing a signal-to-noise ratio of 13 or more is added, the neural network forms a curve of normalised values of modulation-transfer-function against frequency having a higher area thereunder than an area under a corresponding curve for a reference Gaussian smoothing filter. The reference Gaussian smoothing filter is defined with a smoothing variance set such that σ2=4.
The invention also supports computer system implementation and additional constrained behaviors and derived or interpreted imaging outputs as recited in dependent claim form.
Claims Coverage
The independent claims are directed to method and imaging system and computer system implementations that use a neural network filter for diffusion-weighted MRI. Across the independent claims, four core inventive features recur: neural-network-based filtering to improve output signal-to-noise ratio, an MTF-vs-frequency knife-edge evaluation with added Rician noise at SNR≥13 that beats a reference Gaussian smoothing filter, the specification of a Gaussian smoothing variance condition σ2=4 within the evaluation framework, and production of an output image from the acquired image.
Neural network filtering for diffusion-weighted MRI output generation
Using a neural network to filter a diffusion-weighted image of an object acquired by a magnetic resonance imaging scanner, the neural network being programmed to produce an output image from the acquired image.
Improved output signal-to-noise ratio relative to acquired image
The neural network improves a signal to noise ratio of the output image relative to the acquired image.
Knife-edge Rician-noise MTF performance better than reference Gaussian smoothing filter
When applied to a synthetic knife-edge image to which Rician noise providing a signal-to-noise ratio of 13 or more is added, the neural network forms a curve of normalised values of modulation-transfer-function against frequency which has a higher area thereunder than an area under a corresponding curve of normalised values of modulation-transfer-function against frequency for a reference Gaussian smoothing filter.
Gaussian smoothing variance condition for the reference filter
A reference Gaussian smoothing filter relationship is defined with a smoothing variance set such that σ2=4, with a convolution parameter {tilde over (G)}.
Overall, the independent claims cover neural-network filtering of diffusion-weighted MRI images to improve output signal-to-noise ratio, and they require validated performance behavior on a synthetic knife-edge test with Rician noise (SNR≥13) that yields superior modulation-transfer-function curve area compared to a reference Gaussian smoothing filter with σ2=4.
Stated Advantages
Improves a signal to noise ratio of the output image relative to the acquired image.
Produces a higher area under a normalised modulation-transfer-function curve against frequency than a reference Gaussian smoothing filter in the synthetic knife-edge test with Rician noise at SNR of 13 or more.
Documented Applications
Assessing disease extent in a human or animal subject using the output image or an image derived therefrom.
Performing diffusion-weighted MRI using the neural-network filtering to produce an output image from an acquired diffusion-weighted image.
Interested in licensing this patent?