A simple and fast adaptive nonlocal multispectral filtering algorithm for efficient noise reduction in magnetic...

Area of Science:

• Medical Imaging
• Image Processing
• Magnetic Resonance Imaging

Background:

• Multispectral (MS) nonlocal (NL) filters, such as MS-NLML, offer excellent noise reduction in MRI.
• However, MS-NLML's extensive processing time limits its routine clinical application.

Purpose of the Study:

• Introduce a fast, simple, and robust filter, nonlocal estimation of multispectral magnitudes (NESMA), for MS MRI noise reduction.
• Compare NESMA with the existing MS-NLML filter in terms of noise reduction and processing efficiency.
• Introduce adaptive methods to further enhance filtering performance.

Main Methods:

• Compared NESMA and MS-NLML using simulations and in-vivo analyses.
• Developed two adaptive methods (semi-adaptive and fully adaptive) that allow spatial variation of similarity thresholds (R).
• Evaluated adaptive methods using relative Euclidean distance (RED) similarity thresholds and an unbiased criterion.

Main Results:

• NESMA demonstrated comparable filtering performance to MS-NLML.
• NESMA offered significantly faster processing times and simpler implementation.
• Adaptive methods further reduced noise for both NESMA and MS-NLML compared to non-adaptive approaches.

Conclusions:

• NESMA is a fast, robust, and easily implementable filter for MS MRI noise reduction.
• Its efficiency makes it suitable for routine clinical use and analysis of large MRI datasets.
• Adaptive filtering strategies enhance noise reduction capabilities.