Multichannel compressed sensing MR image reconstruction using statistically optimized nonlinear diffusion.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Compressed Sensing

Background:

• Total Variation (TV) based methods are common for compressed-sensing image reconstruction.
• These methods often require parametric tuning, which can be time-consuming and complex.
• Achieving high image quality in multichannel compressed sensing is challenging.

Purpose of the Study:

• To eliminate the need for parametric tuning in total variation (TV) based multichannel compressed-sensing image reconstruction.
• To achieve this using statistically optimized nonlinear diffusion.
• To ensure image quality is not compromised during reconstruction.

Main Methods:

• Employs nonlinear diffusion with a statistically estimated contrast parameter.
• The parameter separates noise and true image edges based on gradient variance.
• Utilizes acquired k-space data to bias the diffusion process towards an optimal solution.

Main Results:

• The proposed method removes the need for extensive parameter tuning required by TV-based methods.
• Statistical estimation of the contrast parameter adapts to varying datasets and undersampling levels.
• Comparable image quality to a-priori tuned TV-based reconstruction was observed.

Conclusions:

• Statistically optimized nonlinear diffusion offers an alternative to TV-based reconstruction without compromising image quality.
• It avoids the time complexity and extensive tuning associated with traditional TV methods.
• This approach is practical for ad-hoc applications in multicoil compressed-sensing reconstruction.