FastPET: Near Real-Time Reconstruction of PET Histo-Image Data Using a Neural Network

Whiteley, William; Panin, Vladimir; Zhou, Chuanyu; Cabello, Jorge; Bharkhada, Deepak; Gregor, Jens

doi:10.1109/trpms.2020.3028364

Cited by 51 publications

(39 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent DL developments in PET image reconstruction can be divided into three groups: (1) direct mapping of PET raw data (i.e. sinograms) to PET images using end-to-end networks [11][12][13]; (2) deep learning reconstruction (DLR), which combines DL with Bayesian reconstruction methods [14]; (3) deep learning enhancement (DLE) of PET images for noise reduction [15,16] or improved convergence [17]. Direct methods aim to learn the whole reconstruction process from scratch; hence, their training is computationally intensive and requires big datasets, whereas DLR methods aim to merge the model-based Bayesian algorithms with CNNs to reduce their data requirement and computational burden.…”

Section: Introductionmentioning

confidence: 99%

Image enhancement of whole-body oncology [18F]-FDG PET scans using deep neural networks to reduce noise

Mehranian

Wollenweber²,

Walker

et al. 2021

Eur J Nucl Med Mol Imaging

View full text Add to dashboard Cite

Purpose To enhance the image quality of oncology [18F]-FDG PET scans acquired in shorter times and reconstructed by faster algorithms using deep neural networks. Methods List-mode data from 277 [18F]-FDG PET/CT scans, from six centres using GE Discovery PET/CT scanners, were split into ¾-, ½- and ¼-duration scans. Full-duration datasets were reconstructed using the convergent block sequential regularised expectation maximisation (BSREM) algorithm. Short-duration datasets were reconstructed with the faster OSEM algorithm. The 277 examinations were divided into training (n = 237), validation (n = 15) and testing (n = 25) sets. Three deep learning enhancement (DLE) models were trained to map full and partial-duration OSEM images into their target full-duration BSREM images. In addition to standardised uptake value (SUV) evaluations in lesions, liver and lungs, two experienced radiologists scored the quality of testing set images and BSREM in a blinded clinical reading (175 series). Results OSEM reconstructions demonstrated up to 22% difference in lesion SUVmax, for different scan durations, compared to full-duration BSREM. Application of the DLE models reduced this difference significantly for full-, ¾- and ½-duration scans, while simultaneously reducing the noise in the liver. The clinical reading showed that the standard DLE model with full- or ¾-duration scans provided an image quality substantially comparable to full-duration scans with BSREM reconstruction, yet in a shorter reconstruction time. Conclusion Deep learning–based image enhancement models may allow a reduction in scan time (or injected activity) by up to 50%, and can decrease reconstruction time to a third, while maintaining image quality.

show abstract

Section: Introductionmentioning

confidence: 99%

Image enhancement of whole-body oncology [18F]-FDG PET scans using deep neural networks to reduce noise

Mehranian

Wollenweber²,

Walker

et al. 2021

Eur J Nucl Med Mol Imaging

View full text Add to dashboard Cite

show abstract

“…Our investigation of fast PET [20], based on Deep Leaning, where the goal is achieving high quality, not necessary diagnostic, PET images from histo-images in real time is currently undergoing. It will be interesting to understand if additional DL processing of histo-images will result in MVF estimation improvement.…”

Section: Discussionmentioning

confidence: 99%

Histo-Projections TOF Data Non-Rigid Motion Estimation and Correction

Panin¹

2021

Preprint

Self Cite

View full text Add to dashboard Cite

<p>With advancements in PET technology, scanners with a very good time-of-flight (TOF) resolution became available. In addition, scanners with long axial extent are now commercialized. Although raw data sizes increase significantly with such scanners, good TOF resolution allows substantial data compression without any loss of spatial resolution: for example, the histo-projections with relatively coarse sampling in the TOF direction in the Siemens Biograph Vision and Vision Quadra PET/CT scanners. The patient motion during the scan is unavoidable. The pattern caused by breathing may result in a relatively large displacement of organs and the consequent blurring of clinically relevant features in regions affected by the motion. As was shown in our previous work, non-rigid motion correction can be performed in the quasi-image space of histo-projections. The TOF locality property can be used to locally perform motion correction; that is, the approximation of motion as locally rigid on a scale of TOF resolution. In this work, we investigate motion estimation in the TOF data space from histo-images by histogramming the data into a 3D array consistent (in geometry and size) with the reconstruction image. The histo-image estimated motion is then used in histo-projections motion correction. Initial results using an XCAT digital phantom showed that the presented methodology accommodates for changes in non-rigid body movements for a typical pattern of patient motion. The patient scan on newly developed Biograph Vision Quadra PET/CT scanner was used in validation of presented motion correction.</p>

show abstract

“…The potential advantage of direct use of raw PET (sinogram) data (whether in direct methods or unrolled methods) perhaps is still in need of more convincing demonstration. Therefore, methods like that of Whiteley et al [77] with their use of TOF backprojected images as the starting point for deep learning, do look promising in the near future.…”

Section: Summary and Future Perspectivesmentioning

confidence: 99%

Deep Learning for PET Image Reconstruction

Reader

Corda-DrIncan

Mehranian

et al. 2021

IEEE Trans. Radiat. Plasma Med. Sci.

170

View full text Add to dashboard Cite

This article reviews the use of a subdiscipline of artificial intelligence (AI), deep learning, for the reconstruction of images in positron emission tomography (PET). Deep learning can be used either directly or as a component of conventional reconstruction, in order to reconstruct images from noisy PET data. The review starts with an overview of conventional PET image reconstruction and then covers the principles of general linear and convolution-based mappings from data to images, and proceeds to consider nonlinearities, as used in convolutional neural networks (CNNs). The direct deep-learning methodology is then reviewed in the context of PET reconstruction. Direct methods learn the imaging physics and statistics from scratch, not relying on a priori knowledge of these models of the data. In contrast, model-based or physics-informed deep-learning uses existing advances in PET image reconstruction, replacing conventional components with deep-learning data-driven alternatives, such as for the regularization. These methods use trusted models of the imaging physics and noise distribution, while relying on training data examples to learn deep mappings for regularization and resolution recovery. After reviewing the main examples of these approaches in the literature, the review finishes with a brief look ahead to future directions.

show abstract

FastPET: Near Real-Time Reconstruction of PET Histo-Image Data Using a Neural Network

Cited by 51 publications

References 46 publications

Image enhancement of whole-body oncology [18F]-FDG PET scans using deep neural networks to reduce noise

Image enhancement of whole-body oncology [18F]-FDG PET scans using deep neural networks to reduce noise

Histo-Projections TOF Data Non-Rigid Motion Estimation and Correction

Deep Learning for PET Image Reconstruction

Contact Info

Product

Resources

About