Deep Neural Network for Selective Streak Artifact Removal in Medical Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical imaging technologies, such as MRI and CT, face challenges in rapidly removing streak artifacts and noise from images acquired with incomplete data, leading to reduced image quality and increased reconstruction time, especially in real-time imaging applications.
Innovation Solution
The use of deep neural networks to selectively and independently remove streak artifacts and noise from medical images by mapping them to residual components, allowing for flexible control over the de-noising process and reducing computational intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If incomplete acquisition of measurement data is used to increase image acquisition speed, then image acquisition speed is improved, but imaging artifacts and reduced signal-to-noise ratio occur
Solution Approach 1:
The patent extracts and removes imaging artifacts from the reconstructed image by processing the image data to identify and eliminate artifact components while preserving the underlying anatomical information. This is achieved through image processing algorithms that separate artifact signals from valid image signals.
Solution Approach 2:
The patent applies parameter changes by adjusting reconstruction parameters, filtering parameters, or processing parameters to reduce artifacts. This may include modifying regularization parameters, filter cutoff frequencies, or iterative reconstruction parameters to optimize the balance between artifact removal and image quality preservation.
2Object-affected harmful factors
If sophisticated reconstruction techniques are used to remove imaging artifacts, then imaging artifacts are reduced, but reconstruction time increases
Solution Approach 1:
The patent applies partial action by selectively removing only the most prominent artifacts or applying artifact removal to specific regions of interest rather than processing the entire image uniformly. This reduces computational burden while maintaining effective artifact reduction in critical areas.
Solution Approach 2:
The patent employs accelerated reconstruction algorithms that skip certain computationally intensive steps or use approximate methods that provide sufficient artifact removal with reduced computation time. This may include using fast Fourier transform optimizations or simplified iterative reconstruction schemes.
3Object-affected harmful factors
If CS reconstruction is used to reduce imaging artifacts, then imaging artifacts are reduced, but computational intensity and reconstruction complexity increase
Solution Approach 1:
The patent employs simpler, computationally less intensive artifact removal methods that can be applied quickly without requiring complex optimization algorithms. These methods may include conventional filtering techniques or simplified iterative approaches that provide adequate artifact reduction with lower computational cost and simpler implementation.
4Object-affected harmful factors
If conventional artifact removal methods are used, then imaging artifacts are reduced, but control over independent removal of different artifact types is limited
Solution Approach 1:
The patent segments the artifact removal process into distinct components, allowing independent processing of different artifact types. By separating the removal of streak artifacts, motion artifacts, and noise into independent processing steps or parameters, the system provides flexible control over the extent and type of artifact removal applied to the image.
Data Source
AI summary
Methods and systems are provided for independently removing streak artifacts and noise from medical images, using trained deep neural networks. In one embodiment, streak artifacts and noise may be selectively and independently removed from a medical image by receiving the medical image comprising streak artifacts and noise, mapping the medical image to a streak residual and a noise residual using the trained deep neural network, subtracting the streak residual from the medical image to a first extent, and subtracting the noise residual from the medical image to a second extent, to produce a de-noised medical image, and displaying the de-noised medical image via a display device.