• Corpus ID: 239616237

A Data-Driven Reconstruction Technique based on Newton's Method for Emission Tomography

  title={A Data-Driven Reconstruction Technique based on Newton's Method for Emission Tomography},
  author={Loizos Koutsantonis and Tiago Carneiro and Emmanuel Kieffer and Fr{\'e}d{\'e}ric Pinel and Pascal Bouvry},
In this work, we present the Deep Newton Reconstruction Network (DNR-Net), a hybrid data-driven reconstruction technique for emission tomography inspired by Newton’s method, a well-known iterative optimization algorithm. The DNR-Net employs prior information about the tomographic problem provided by the projection operator while utilizing deep learning approaches to a) imitate Newton’s method by approximating the Newton descent direction and b) provide data-driven regularisation. We demonstrate… 

Figures and Tables from this paper


Fast Quasi-Newton Algorithms for Penalized Reconstruction in Emission Tomography and Further Improvements via Preconditioning
This paper reports on the feasibility of using a quasi-Newton optimization algorithm, limited-memory Broyden-Fletcher-Goldfarb-Shanno with boundary constraints (L-BFGS-B), for penalized image
Deep learning-based image reconstruction for TOF PET with DIRECT data partitioning format
A novel deep learning-based image reconstruction method for TOF PET with the help of the DIRECT data partitioning method, no fully connected layers were used and 3D image reconstruction can be directly achieved within the limits of the current hardware.
Accelerating image reconstruction in ultrasound transmission tomography using L-BFGS algorithm
The limited memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) algorithm is investigated as a preconditioning technique due to its ability to efficiently approximate inverse Hessian without performing forward model or its adjoint, and is competitive with the accelerated CG based methods regarding the number of iterations, and outperforms them regarding reconstruction time.
Few-view single photon emission computed tomography (SPECT) reconstruction based on a blurred piecewise constant object model.
The results suggest that the algorithm may provide high reconstruction accuracy even when the object does not match the assumed blurring model, and introduce patchy artifacts in some reconstructed images, depending on the noise level and the selected algorithm parameters.
A penalized-likelihood image reconstruction method for emission tomography, compared to postsmoothed maximum-likelihood with matched spatial resolution
This paper performed positron emission tomography and single photon emission computed tomography simulations to compare the performance of the new method to that of the postsmoothed maximum-likelihood (ML) approach, using the impulse response of the former method as the post smoothing filter for the latter.
DirectPET: full-size neural network PET reconstruction from sinogram data
The ability of DirectPET to quickly reconstruct high-quality, multislice image volumes suggests potential clinical viability of the method, but design parameters and performance boundaries need to be fully established before adoption can be considered.
A Learned Reconstruction Network for SPECT Imaging
A neural network designed specifically for single photon emission computed tomography (SPECT) image reconstruction was developed and it was shown that the network still provided sharp images when using reduced view projection data (retrained with reduced view data).
A Reconstruction Method Based on Deep Convolutional Neural Network for SPECT Imaging
The results show that the proposed CNNR model outperforms both FBP and MLEM methods.
Improved Low-Count Quantitative PET Reconstruction With an Iterative Neural Network
The proposed BCD-Net significantly improves CNR and RMSE of the reconstructed images compared to MBIR methods using non-trained regularizers, total variation (TV) and non-local means (NLM), and generalizes to test data that differs from the training data.