KW Filtering for Dynamic MR Image De-noising
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Real-time cardiac magnetic resonance (MR) cine imaging faces challenges in maintaining signal-to-noise ratio (SNR) due to noise corruption during acquisition and reconstruction, which affects image quality and interpretation, particularly in dynamic MR image series where existing de-noising techniques either sacrifice image sharpness or fail to effectively utilize long-range temporal correlations.
Innovation Solution
The Karhunen-Loeve Transform (KLT) is combined with Wavelet filtering to de-noise dynamic MR image series, utilizing 1D temporal KLT to compress signal information and discard noise-only frames, followed by adaptive Wavelet filtering to reduce noise while preserving important signal information, thereby achieving significant noise reduction with minimal loss in image sharpness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If spatial filtering (e.g., Wavelet filtering) is applied to remove noise, then noise is reduced, but edge sharpness and fine structures are blurred
Solution Approach 1:
The patent combines temporal filtering (KLT) with spatial filtering (Wavelet) to create a hybrid approach. The KLT component exploits temporal correlations across the image series to remove noise, while the Wavelet component handles spatial de-noising. This combination allows each filter to work on its strength, reducing the need for aggressive spatial filtering that would blur edges.
Solution Approach 2:
The KLT filtering is applied first to pre-process the image series by removing noise that can be eliminated through temporal correlation. This preliminary de-noising step reduces the noise burden before the spatial Wavelet filtering stage, allowing the Wavelet filter to operate more conservatively and preserve edge sharpness.
2Object-affected harmful factors
If aggressive thresholding is applied in Wavelet de-noising to maximize SNR gain, then noise is reduced more effectively, but image blurring increases
Solution Approach 1:
The patent merges two filtering approaches with different characteristics: KLT which is effective for temporally correlated noise, and Wavelet which preserves spatial features. By combining them, the system achieves effective noise removal without relying on aggressive single-threshold Wavelet filtering that causes blurring.
Solution Approach 2:
The patent applies different filtering strategies to different components of the image processing pipeline. KLT is applied to exploit temporal correlations, while Wavelet is applied for spatial de-noising. Each component is optimized for its specific function, allowing local optimization of noise removal versus detail preservation.
3Speed
If parallel MRI acceleration is applied to reduce scan time, then temporal resolution is improved, but SNR reduces by the square root of the acceleration factor
Solution Approach 1:
The KLT method copies information across multiple frames in the image series, exploiting the temporal redundancy and quasi-periodic nature of cardiac motion. By integrating information from multiple time points, the effective SNR is improved without requiring additional physical signal averaging that would increase scan time.
Solution Approach 2:
The hybrid KLT-Wavelet filtering combines temporal and spatial filtering mechanisms to address the SNR penalty from parallel MRI acceleration. The KLT component recovers SNR through temporal correlation, compensating for the acceleration-induced noise increase while maintaining the fast temporal resolution.
4Ease of operation
If single-threshold Wavelet de-noising is applied to simplify the process, then ease of operation is improved, but there is a trade-off between signal loss and SNR gain
Solution Approach 1:
The patent combines two filtering methods with different strengths: KLT for temporal de-noising and Wavelet for spatial de-noising. This combination allows the system to achieve better overall performance than either method alone, balancing simplicity with effectiveness.
Solution Approach 2:
Instead of using a single aggressive threshold, the patent applies a more conservative approach by first using KLT to remove easily eliminatable noise, then applying moderate Wavelet thresholding. This partial action at each stage avoids the need for excessive thresholding that would cause signal loss.
Data Source
AI summary
A hybrid filtering method called Karhunen Loeve Transform-Wavelet (KW) filtering is presented to de-noise dynamic cardiac magnetic resonance images that simultaneously takes advantage of the intrinsic spatial and temporal redundancies of real-time cardiac cine. This filtering technique combines a temporal Karhunen-Loeve transform (KLT) and spatial adaptive wavelet filtering. KW filtering has four steps. The first is applying the KLT along the temporal direction, generating a series of “eigenimages”. The second is applying Marcenko-Pastur (MP) law to identify and discard the noise-only eigenimages. The third applying a 2-D spatial wavelet filter with adaptive threshold to each eigenimage to define the wavelet filter strength for each of the eigenimages based on the noise variance and standard deviation of the signal. Lastly, the inverse KLT is applied to the filtered eigenimages to generate a new series of cine images with reduced image noise.