Interval-based reconstruction for uncertainty quantification in PET

Kucharczak, Florentin; Loquin, Kevin; Buvat, Irène; Strauss, Olivier; Mariano‐Goulart, Denis

doi:10.1088/1361-6560/aa9ea6

Cited by 11 publications

(7 citation statements)

References 33 publications

(61 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The potential of applying Efron’s statistical bootstrap ( Efron and Tibshirani, 1994 ) in this setting was described by ( Haynor and Woods, 1989 ). There have been a number of subsequent contributions - see, for example, ( Buvat, 2002 ; Dahlbom, 2002 ; Lartizien et al, 2010 ; Ibaraki et al, 2014 ; Kucharczak et al, 2018 ) - that have attempted to implement variations on this approach. The attraction of the non-parametric bootstrap is that it does not involve detailed analytic assumptions which may be difficult to justify in a real patient study.…”

Section: Introductionmentioning

confidence: 99%

A Generalized Linear modeling approach to bootstrapping multi-frame PET image data

O’Sullivan

et al. 2021

Medical Image Analysis

View full text Add to dashboard Cite

PET imaging is an important diagnostic tool for management of patients with cancer and other diseases. Medical decisions based on quantitative PET information could potentially benefit from the availability of tools for evaluation of associated uncertainties. Raw PET data can be viewed as a sample from an inhomogeneous Poisson process so there is the possibility to directly apply bootstrapping to raw projection-domain list-mode data. Unfortunately this is computationally impractical, particularly if data reconstruction is iterative or the acquisition protocol is dynamic. We develop a flexible statistical linear model analysis to be used with multi-frame PET image data to create valid bootstrap samples. The technique is illustrated using data from dynamic PET studies with fluoro-deoxyglucose (FDG) and fluoro-thymidine (FLT) in brain and breast cancer patients. As is often the case with dynamic PET studies, data have been archived without raw list-mode information. Using the bootstrapping technique maps of kinetic parameters and associated uncertainties are obtained. The quantitative performance of the approach is assessed by simulation. The proposed image-domain bootstrap is found to substantially match the projection-domain alternative. Analysis of results points to a close relation between relative uncertainty in voxel-level kinetic parameters and local reconstruction error. This is consistent with statistical theory.

show abstract

Section: Introductionmentioning

confidence: 99%

A Generalized Linear modeling approach to bootstrapping multi-frame PET image data

O’Sullivan

et al. 2021

Medical Image Analysis

View full text Add to dashboard Cite

show abstract

“…Images were sampled on a 400x400x109 grid with a voxel size of 2.04x2.04x2.03 mm 3 . Second, PET data were reconstructed using an interval-valued extension of the maximum likelihood-expectation maximization (ML-EM) algorithm (33,34) called NIBEM (29,30) that stands for Non-additive Interval Based Expectation Maximization. The main motivation for using this algorithm resides in its ability to directly reconstruct voxel-wise confidence intervals.…”

Section: Pet Data Acquisition and Reconstructionmentioning

confidence: 99%

“…The considered confidence intervals account for the statistical variability affecting reconstructed voxel values. The confidence level associated to these intervals was shown to be about 90% (29). As the current version of the mentioned algorithm was only described in 2D, 3D emission data were rebinned into a stack of 109 2-dimensional (2D) sinograms using the Fourier rebinning (FORE) algorithm (35).…”

Section: Pet Data Acquisition and Reconstructionmentioning

confidence: 99%

“…The aim of the present study is to propose and highlight the potential and the relevance of brain 18 F-FDG PET analysis based on an interval-valued reconstruction of 18 F-FDG distribution (29,30). The proposed framework allows for a direct estimation of voxel-wise confidence intervals accounting for the statistical variability of voxel values.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Brain 18F-FDG PET analysis via interval-valued reconstruction: proof of concept for Alzheimer’s disease diagnosis

et al. 2020

Self Cite

View full text Add to dashboard Cite

Objective: We propose an innovative approach for 18 F-FDG PET analysis based on an interval-valued reconstruction of 18 F-FDG brain distribution. Its diagnostic performance for Alzheimer's disease (AD) diagnosis with comparison to a validated post-processing software was assessed.Method: Brain 18 F-FDG PET data from 26 subjects were acquired in a clinical routine setting. Raw data were reconstructed using an interval-valued version of the ML-EM algorithm called NIBEM that stands for Non-additive Interval Based Expectation Maximization. Subject classification was obtained via interval-based statistical comparison (intersection ratio, IR) between cortical regions of interest (ROI) including parietal, temporal, and temporo-mesial cortices and a reference region, the sub-cortical grey nuclei, known not to be affected by AD. In parallel, PET images were post-processed using a validated automated software based on the computation of ROI normalized uptake ratios standard deviation (SUVr SD) with reference to a healthy control database (Siemens Scenium). Clinical diagnosis made during follow-up was considered as the gold-standard for patient classification (16 healthy controls and 10 AD patients).Results: Both methods provided cortical ROI indices that were significantly different between controls and AD patients. The area under the ROC curve for control/AD classification was statistically identical (0.96 for NIBEM IR and 0.95 for Scenium SUVr SD). At the optimal threshold, the sensitivity, specificity, accuracy, positive predictive value and negative predictive value were respectively 100%, 88%, 92%, 83%, and 100% for both Scenium SUVr SD and NIBEM IR methods. Conclusion:This preliminary study shows that interval-valued reconstruction allows selfconsistent analysis of brain 18 F-FDG PET data, yielding diagnostic performances that seems promising with respect to those of a commercial post-processing software based on SUVr SD analysis.

show abstract

“…Several methods for estimating PET image noise (and hence uncertainty) have been proposed, such as the interval-based image reconstruction Kucharczak et al. (2018) , Bayesian estimation Sitek (2012) , approximate variance estimation of regularised reconstruction methods Fessler (1996) ; Jinyi Qi and Leahy (2000) and variance estimation of the expectation-maximisation (EM) algorithm used in PET image reconstruction Barrett et al.…”

Section: Introductionmentioning

confidence: 99%

Uncertainty analysis of MR-PET image registration for precision neuro-PET imaging

et al. 2021

View full text Add to dashboard Cite

Highlights Novel methodology and software for MR-PET registration uncertainty analysis. Registration software had the biggest effect on MR-PET registration precision, followed by reconstruction parameters (i.e., iterations, smoothing) and PET count level. PVC can significantly improve the PET signal, but since it relies on precise MR-PET registration, it also increases PET signal variability and hence care should be taken when using it.

show abstract

Interval-based reconstruction for uncertainty quantification in PET

Cited by 11 publications

References 33 publications

A Generalized Linear modeling approach to bootstrapping multi-frame PET image data

A Generalized Linear modeling approach to bootstrapping multi-frame PET image data

Brain 18F-FDG PET analysis via interval-valued reconstruction: proof of concept for Alzheimer’s disease diagnosis

Uncertainty analysis of MR-PET image registration for precision neuro-PET imaging

Contact Info

Product

Resources

About