Nonparametric Residue Analysis of Dynamic PET Data With Application to Cerebral FDG Studies in Normals

O’Sullivan, Finbarr; Muzi, Mark; Spence, Alexander M.; Mankoff, David; O’Sullivan, Janet; Fitzgerald, Niall; Newman, George C.; Krohn, Kenneth A.

doi:10.1198/jasa.2009.0021

Cited by 29 publications

(62 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The scaling factor in Equation (9) differs from the one reported in the paper by O'Sullivan et al 3 We closely checked our derivation of the scaling factor under the assumption that the residence density is exponential (h(t) = α × exp(− αt)). We confirmed our derivation using the reported clinical and simulated data.…”

Section: Outcome Measuresmentioning

confidence: 89%

“…However, according to the theory of intravascular indicator dilutions, which was developed in the seminal work of Meier and Zierler 22 and then extended to the general case of extravascular tracers (as in the case of PET) by O'Sullivan et al, 3 an equation analog to (1) can be written:…”

Section: Interpretation Of Positron Emission Tomography Datamentioning

confidence: 99%

“…3 However, once the IRF has been nonparametrically estimated via deconvolution, it remains to establish which are the outcomes that can be actually computed from the IRF, and how they compare with respect to traditional outcomes used in PET. The first outcome measure usually computed in PET for a reversible tracer is V T , whose formulation in terms of the 2TCM K 1 to k 4 values is: 2…”

Section: Outcome Measuresmentioning

confidence: 99%

“…Nevertheless, they have been rarely adopted in PET. O'Sullivan et al 3 Several approaches have been developed that fall between these two extremes. [5][6][7] Generally, these involve expressing the IRF in terms of basis functions that, although derived from the principles of linear time-invariant systems and based on CMs, 8 do not impose any predefined compartmental structure.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Model-Free Quantification of Dynamic PET Data Using Nonparametric Deconvolution

Zanderigo

Parsey

Ogden

2015

J Cereb Blood Flow Metab

View full text Add to dashboard Cite

Dynamic positron emission tomography (PET) data are usually quantified using compartment models (CMs) or derived graphical approaches. Often, however, CMs either do not properly describe the tracer kinetics, or are not identifiable, leading to nonphysiologic estimates of the tracer binding. The PET data are modeled as the convolution of the metabolite-corrected input function and the tracer impulse response function (IRF) in the tissue. Using nonparametric deconvolution methods, it is possible to obtain model-free estimates of the IRF, from which functionals related to tracer volume of distribution and binding may be computed, but this approach has rarely been applied in PET. Here, we apply nonparametric deconvolution using singular value decomposition to simulated and test-retest clinical PET data with four reversible tracers well characterized by CMs (

show abstract

Section: Outcome Measuresmentioning

confidence: 89%

Section: Interpretation Of Positron Emission Tomography Datamentioning

confidence: 99%

Section: Outcome Measuresmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Model-Free Quantification of Dynamic PET Data Using Nonparametric Deconvolution

Zanderigo

Parsey

Ogden

2015

J Cereb Blood Flow Metab

View full text Add to dashboard Cite

show abstract

“…This assumption is at odds with micro-vascular flow measurements which support a more log-Normal or Gamma-like form for the transit time density. If the residue is to be estimated nonparametrically, it is desirable to have a procedure, like that given in Hawe et al (2012) or O’Sullivan et al (2009), that does not impose an unrealistic physiologic assumption on the residue function ab initio.…”

Section: Introductionmentioning

confidence: 99%

Voxel-level mapping of tracer kinetics in PET studies: A statistical approach emphasizing tissue life tables

et al. 2014

Self Cite

View full text Add to dashboard Cite

Most radiotracers used in dynamic positron emission tomography (PET) scanning act in a linear time-invariant fashion so that the measured time-course data are a convolution between the time course of the tracer in the arterial supply and the local tissue impulse response, known as the tissue residue function. In statistical terms the residue is a life table for the transit time of injected radiotracer atoms. The residue provides a description of the tracer kinetic information measurable by a dynamic PET scan. Decomposition of the residue function allows separation of rapid vascular kinetics from slower blood-tissue exchanges and tissue retention. For voxel-level analysis, we propose that residues be modeled by mixtures of nonparametrically derived basis residues obtained by segmentation of the full data volume. Spatial and temporal aspects of diagnostics associated with voxel-level model fitting are emphasized. Illustrative examples, some involving cancer imaging studies, are presented. Data from cerebral PET scanning with 18F fluoro-deoxyglucose (FDG) and 15O water (H2O) in normal subjects is used to evaluate the approach. Cross-validation is used to make regional comparisons between residues estimated using adaptive mixture models with more conventional compartmental modeling techniques. Simulations studies are used to theoretically examine mean square error performance and to explore the benefit of voxel-level analysis when the primary interest is a statistical summary of regional kinetics. The work highlights the contribution that multivariate analysis tools and life-table concepts can make in the recovery of local metabolic information from dynamic PET studies, particularly ones in which the assumptions of compartmental-like models, with residues that are sums of exponentials, might not be certain.

show abstract

Total‐body parametric imaging using the Patlak model: Feasibility of reduced scan time

Feng

Shen

et al. 2022

Medical Physics

View full text Add to dashboard Cite

Purpose This study explored the feasibility of reducing the scan time of Patlak parametric imaging on the uEXPLORER. Methods A total of 65 patients (27 females and 38 males, age 56.1 ± 10.4) were recruited in this study. 18F fluorodeoxyglucose was injected, and its dose was adjusted by body weight (4.07 MBq/kg). Total‐body dynamic scanning was performed on the uEXPLORER total‐body Positron emission tomography/computed tomography (CT) scanner with a total scan time of 60 min from the injection. The image derived input function (IDIF) was obtained from the aortic arch. The voxelwise Patlak analysis was applied to generate the Ki images designated as GIDIF with different acquisition times (20–60, 30–60, 40‐60, and 44–60 min). The population‐based input function (PBIF) was constructed from the mean value of the IDIF from the population, and Ki images designated as GPBIF were generated using the PBIF. Nonlocalmeans (NLM) denoising was applied to the generated images to get two extra groups of (NLM‐designated) images: GIDIF+NLM and GPBIF+NLM. Two radiologists evaluated the overall image quality, noise, and lesion detectability of the Ki images from different groups. The 20–60 min scans in GIDIF were selected as the gold standard for each patient. We determined that image quality is at sufficient level if all the lesions can be recognized and meet the clinical criteria. Ki values in muscle and lesion were compared across different groups to evaluate the quantitative accuracy. Results The overall image quality, image noise, and lesion conspicuity were significantly better in long time series than short time series in all four groups (all p < 0.001). The Ki images in the GIDIF and GPBIF groups generated from 30‐min scans showed diagnostic value equivalent to the 40‐min scans of GIDIF. While the image quality of the 16‐min scans was poor, all lesions could still be detected. No significant difference was found between Ki values estimated with GIDIF and GPBIF in muscle and lesion regions (all p > 0.5). After applying the NLM filter, the coefficient of variation could be reduced on the order of (1%, 15%, 19%, and 37%) and (110%, 125%, 94%, and 69%) with four acquisition time schemes for lesion and muscle. The reduction percentage did not have a substantial difference in IDIF and PBIF group. The Ki images in the GIDIF+NLM and GPBIF+NLM groups generated from the 20‐min acquisitions showed acceptable quality. All lesions could be found on the NLM processed images of the 16‐min scans. No significant difference was found between Ki values produced with GIDIF+NLM and GPBIF+NLM in muscle and lesion regions(all p > 0.7). Conclusions The Ki images generated by the PBIF‐based Patlak model using a 20‐min dynamic scan with the NLM filter achieved a similar diagnostic efficiency to images with GIDIF from 40‐min dynamic data, and there is no significant difference between Ki images generated using IDIF or PBIF (p > 0.5).

show abstract

Nonparametric Residue Analysis of Dynamic PET Data With Application to Cerebral FDG Studies in Normals

Cited by 29 publications

References 42 publications

Model-Free Quantification of Dynamic PET Data Using Nonparametric Deconvolution

Model-Free Quantification of Dynamic PET Data Using Nonparametric Deconvolution

Voxel-level mapping of tracer kinetics in PET studies: A statistical approach emphasizing tissue life tables

Total‐body parametric imaging using the Patlak model: Feasibility of reduced scan time

Contact Info

Product

Resources

About