MRI Denoising Using Pixel-Wise Threshold Selection

Area of Science:

• Medical Imaging
• Image Processing
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) offers non-invasive medical visualization.
• A key challenge in MRI is balancing image quality with acquisition time.
• Existing denoising methods often cause image distortion or inadequate noise removal.

Purpose of the Study:

• To introduce a novel pixel-wise denoising technique for Magnetic Resonance (MR) images.
• To address the limitations of global thresholding in current MRI denoising algorithms.
• To improve both the visual quality and acquisition speed of MRI scans.

Main Methods:

• A novel pixel-wise thresholding approach using singular vectors from singular value decomposition (SVD).
• Localized singular values are used as thresholds at each pixel to account for spatial noise variations.
• Validation performed on MR images of a standard phantom approved by the Magnetic Resonance Accreditation Program (MRAP).

Main Results:

• Denoised MR images exhibit superior visual quality and recover suppressed structural information.
• Significant improvements observed: Peak-Signal-to-Noise-Ratio (PSNR) increased by ≥ 18%, Contrast-to-Noise-Ratio (CNR) by ≥ 200%.
• Structural Similarity Index (SSIM) of ≥ 0.95 indicates effective recovery of crucial structural information.
• The method demonstrated superior performance compared to standard MRI denoising filters.
• Scan time was reduced by 2-3 times.

Conclusions:

• The pixel-wise denoising technique effectively removes noise and enhances image quality in MRI.
• The method preserves critical structural details, outperforming conventional approaches.
• This technique has the potential for integration into MRI systems to enable faster, higher-quality imaging.