A Blind Deconvolution Technique Based on Projection Onto Convex Sets for Magnetic Particle Imaging

Onur Yorulmaz, Ömer Burak Demirel, Yavuz Muslu, Tolga Çukur, Emine Ülkü Sarıtaş, Ahmet Enis ÇetinarXiv 2017 · DOI · arXiv

Abstract

Magnetic Particle Imaging (MPI) is a promising medical imaging modality that tracks superparamagnetic iron oxide (SPIO) nanoparticles with high contrast and no ionizing radiation. However, the standard "x-space" reconstruction method produces inherently blurry images. This paper introduces a blind deconvolution algorithm for MPI that restores these images without requiring prior knowledge of the system's point spread function (PSF). By exploiting the unique "zero-phase" property of the MPI PSF in the Fourier domain, the authors use a Projections Onto Convex Sets (POCS) approach to significantly improve image resolution and quality, outperforming standard methods that require an explicit PSF model.

Introduction

Deconvolution is essential in MPI to counteract the blurring effects inherent in x-space reconstruction. Traditionally, this requires knowing the system's PSF, which varies based on nanoparticle type and scanner settings. Measuring the PSF for every scenario is impractical. "Blind" deconvolution—recovering the image without knowing the PSF—is an ill-posed problem. This work leverages physical properties of the MPI signal to constrain the solution space effectively.

The Zero-Phase Property

The core insight of this work is that the MPI PSF is essentially the derivative of a symmetric Langevin function. In the Fourier domain, a symmetric spatial function results in a purely real transfer function with zero phase.

1D PSF and its corresponding Transfer Function in the Fourier domain. The transfer function is purely real, indicating zero phase.

This property is robust; even though the reconstructed images are highly blurred, the phase information remains unaltered. This provides a powerful "ground truth" anchor for reconstruction.

Methods: Projections Onto Convex Sets (POCS)

The proposed algorithm iteratively projects the image onto three convex sets to enforce consistency:

  1. Fourier Phase Constraint: Forces the image to match the phase of the original blurry data.
  2. Spatial Support Constraint: Limits the image to the known Field of View (FOV).
  3. $\ell_1$ Norm Regularization: Enforces sparsity and suppresses noise.
2D PSF and its Fourier components. The imaginary part of the Transfer Function (c) is zero, confirming the zero-phase property in two dimensions.

To ensure efficiency, the authors implemented a fast approximate projection for the $\ell_1$ norm and a "boundary stretching" technique to mitigate artifacts from the finite FOV.

Results and Discussion

The algorithm was validated using simulated "vein" phantoms and experimental data from an in-house FFP MPI scanner. In simulations, the POCS method achieved a Peak Signal-to-Noise Ratio (PSNR) 5.14 dB higher than Wiener deconvolution and 2.77 dB higher than the Lucy-Richardson method on average.

Experimentally, the algorithm reduced the apparent width (FWHM) of 2mm vials from 8mm in the raw image to approximately 5mm. Crucially, the method is nanoparticle-agnostic. It successfully deblurred a phantom containing three different nanoparticle environments simultaneously, proving its robustness against model mismatch that often plagues "informed" deconvolution methods.

Conclusion

By leveraging the zero-phase property of the MPI PSF, this blind deconvolution technique provides a robust, high-performance solution for image sharpening without the need for tedious PSF measurements. While challenges remain regarding horizontal resolution and initialization sensitivity, these results represent a significant step toward high-resolution, tracer-independent MPI imaging.

Summary updated

Abstract

Magnetic Particle Imaging (MPI) is an emerging imaging modality that maps the spatial distribution of magnetic nanoparticles. The x-space reconstruction in MPI results in highly blurry images, where the resolution depends on both system parameters and nanoparticle type. Previous techniques to counteract this blurring rely on the knowledge of the imaging point spread function (PSF), which may not be available or may require additional measurements. This work proposes a blind deconvolution algorithm for MPI to recover the precise spatial distribution of nanoparticles. The proposed algorithm exploits the observation that the imaging PSF in MPI has zero phase in Fourier domain. Thus, even though the reconstructed images are highly blurred, phase remains unaltered. We leverage this powerful property to iteratively enforce consistency of phase and bounded l1 energy information, using an orthogonal Projections Onto Convex Sets (POCS) algorithm. To demonstrate the method, comprehensive simulations were performed without and with nanoparticle relaxation effects, and at various noise levels. In addition, imaging experiments were performed on an in-house MPI scanner using a three-vial phantom that contained different nanoparticle types. Image quality was compared with conventional deconvolution methods, Wiener deconvolution and Lucy-Richardson method, which explicitly rely on the knowledge of PSF. Both the simulation results and experimental imaging results show that the proposed blind deconvolution algorithm outperforms the conventional deconvolution methods. Without utilizing the imaging PSF, the proposed algorithm improves image quality and resolution even in the case of different nanoparticle types, while displaying reliable performance against loss of the fundamental harmonic, nanoparticle relaxation effects, and noise.
BibTeX
@misc{yorulmaz2017blind,
  author = {Onur Yorulmaz and Ömer Burak Demirel and Yavuz Muslu and Tolga Çukur and Emine Ülkü Sarıtaş and Ahmet Enis Çetin},
  title = {{A Blind Deconvolution Technique Based on Projection Onto Convex Sets for Magnetic Particle Imaging}},
  year = {2017},
  eprint = {1705.07506},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1705.07506},
}
Figure