Abstract:The realization of fast and high-quality three-dimensional particle-field image characterization is always highly desired in the areas, such as experimental fluid mechanics and biomedicine, etc., as the micro-particle distribution status in a flow-field can characterize the field properties well. In the particle-field image reconstruction and characterization, a popularly-used approach at present is the computed tomography. The great advantage of the computed tomography for particle-field image reconstruction lies in that the full particle spatial distribution can be obtained and presented due to multi-angle sampling. Recently, with the development and application of deep learning techniques in the computed tomography, the image quality is greatly improved by means of the powerful learning ability of a deep learning network. In addition, the deep learning application also makes it possible to speed up the computed tomographic imaging process from sparse-sampling due to the strong image feature extraction ability of the network. However, sparse-sampling would lead to insufficient acquirement of the object information during sampling for the computed tomography. Therefore, a sort of artifact noise would emerge and accompany with the reconstructed images, and thus severely affect the image quality. As there is no universal network approach that can be applicable to all types of objects in the suppression of artifact noise, it is still a challenge in removing the sparse-sampling-induced artifact noise in the computed tomography by now. Therefore, we propose and develop a specific lightweight residual and enhanced convergence neural network (LREC-net) approach for suppressing the artifact noise in the particle-field computed tomography here. In this method, the network input dataset is also optimized in signal-to-noise ratio (SNR) to reduce the input noise and ensure the effective particle image feature extraction of the network during the imaging process. In the LREC-net architecture design, a five layers of lightweight and dual-residual down-sampling are constructed on the basis of typical U-net and Resnet50 to make the LREC-net to be more suitable for the particle-field image reconstruction. Moreover, a fast feature convergence module for rapid particle-field feature acquirement is added to up-sampling process of the network to further promote the network processing efficiency. Apart from the design of LREC-net network itself, the optimization of network input dataset in SNR of images is achieved by finding a fit image reconstruction algorithm that can produce higher-SNR particle images in the computed tomography. This achievement reduces the input noise as much as possible and ensure effective particle-field feature extraction by the network. The simulation analysis and experimental test verify effectiveness of the proposed LREC-net method, which involve the evaluations of SNR changes of the input-output images through the network, the proportion of residual artifact noise as ghost-particles (GPP) in the reconstructed images, and the valid-particle loss proportion (PLP). In contrast to the performances of U-net and Resnet50 under the same imaging conditions, all the data in SNR, GPP and PLP show the great improvement of the image quality due to the application of LREC-net method. Meanwhile, the designed LREC-net method also enhances the network running efficiency to a large extent due to the remarkable reduction of training time. Therefore, this work provides a new and effective approach for developing sparse-sampling-based fast and high-quality particle-field computed tomography.

RETNet: Resolution enhancement Transformer network for magnetic particle imaging based on X-space

Transformer for low concentration image denoising in magnetic particle imaging

Deep learning for improving the spatial resolution of magnetic particle imaging

Dual-channel end-to-end network with prior knowledge embedding for improving spatial resolution of magnetic particle imaging

PET Image Reconstruction with Parallax Correction Based on a Distance-Driven Deep Neural Network

Fusion Networks of CNN and Transformer with Channel Attention for Accurate Tumor Imaging in Magnetic Particle Imaging

PGNet: Projection Generative Network for Sparse-View Reconstruction of Projection-Based Magnetic Particle Imaging.

Image Reconstruction for Magnetic Particle Imaging Based on Sparse Representation and Deep Learning

RIRGAN: an End-to-end Lightweight Multi-Task Learning Method for Brain MRI Super-Resolution and Denoising

MPIGAN: An end-to-end deep based generative framework for high-resolution magnetic particle imaging reconstruction

Spatial-Frequency Multi-Scale Transformer for Deblurring and Shape-Preserving Reconstruction in Magnetic Particle Imaging

TRANS-Net: Transformer-Enhanced Residual-Error AlterNative Suppression Network for MRI Reconstruction.

Artifact noise suppression of particle-field computed tomography based on lightweight residual and enhanced convergence neural network

Magnetic resonance image denoising for Rician noise using a novel hybrid transformer‐CNN network (HTC‐net) and self‐supervised pretraining

Research of Magnetic Particle Imaging Reconstruction Based on the Elastic Net Regularization.

Cross-domain knowledge transfer based parallel-cascaded multi-scale attention network for limited view reconstruction in projection magnetic particle imaging

Magnetic Resonance Image Processing Transformer for General Reconstruction

Super-resolved Reconstruction Method for Spatiotemporally Encoded Magnetic Resonance Imaging Based on Deep Neural Network

Anisotropic edge-preserving network for resolution enhancement in unidirectional Cartesian magnetic particle imaging

Brain MRI Super-Resolution Using Coupled-Projection Residual Network

Super-resolution Magnetic Resonance Imaging Reconstruction Using Deep Attention Networks