Comparison of simplified quantitative analyses of FDG uptake

https://doi.org/10.1016/S0969-8051(00)00143-8Get rights and content

Abstract

Quantitative analysis of [18F]-fluoro-deoxyglucose (FDG) uptake is important in oncologic positron emission tomography (PET) studies to be able to set an objective threshold in determining if a tissue is malignant or benign, in assessing response to therapy, and in attempting to predict the aggressiveness of an individual tumor. The most common method used today for simple, clinical quantitation is standardized uptake value (SUV). SUV is normalized for body weight. Other potential normalization factors are lean body mass (LBM) or body surface area (BSA). More complex quantitation schemes include simplified kinetic analysis (SKA), Patlak graphical analysis (PGA), and parameter optimization of the complete kinetic model to determine FDG metabolic rate (FDGMR). These various methods were compared in a group of 40 patients with colon cancer metastatic to the liver. The methods were assessed by (1) correlation with FDGMR, (2) ability to predict survival using Kaplan-Meier plots, and (3) area under receiver operating characteristic (ROC) curves for distinguishing between tumor and normal liver. The best normalization scheme appears to be BSA with minor differences depending on the specific formula used to calculate BSA. Overall, PGA is the best predictor of outcome and best discriminator between normal tissue and tumor. SKA is almost as good. In conventional PET imaging it is worthwhile to normalize SUV using BSA. If a single blood sample is available, it is possible to use the SKA method, which is distinctly better. If more than one image is available, along with at least one blood sample, PGA is feasible and should produce the most accurate results.

Introduction

Positron emission tomography (PET) is becoming increasingly useful in the field of nuclear medicine. A major advantage of PET over conventional gamma camera imaging is that tissue activity can be accurately determined. Numerous investigators have taken advantage of this capability to use PET imaging to determine blood flow, glucose metabolism, oxygen metabolism, receptor affinity, and other physiologic parameters. Almost all these determinations require blood time activity data in addition to tissue time activity data obtained from multiple, sequential PET images. Such studies require considerable time and manpower to complete, but are necessary in research efforts addressing specific problems that can not be approached with any other noninvasive method.

The primary radiopharmaceutical used in clinical PET studies is [18F]-fluoro-deoxyglucose (FDG), and the most common type of study is an examination of the patient with cancer. The goal of such a study is usually to determine if tumor is present or to determine if a known tumor is responding to treatment. In both types of studies quantitation of tumor FDG metabolism is important. A metabolic rate threshold can be defined to determine whether or not a tissue is malignant. A decrease in FDG metabolism can be used as an indicator that a tumor is responding to therapy and vice versa. In the daily clinical use of PET, because of time and cost considerations, it is usually not feasible to undertake the complex protocols and data analysis used in research studies, and therefore simpler methods must be used.

The quantitative method used most commonly in clinical FDG PET imaging is the standard uptake value (SUV), which is the ratio of activity in a tissue (in μCi/mL) divided by the decay-corrected activity injected into the patient (in μCi/g). The resultant number is almost unitless (actually g/mL) and is a crude measure of degree of uptake of FDG into any tissue. There are numerous problems with the use of SUV, discussed in detail by Hamberg et al. (6) and Keyes (10). Several correction schemes, involving blood glucose level, body surface area, and lean body mass have been proposed to try to improve the accuracy of the SUV method and are now being used at some PET centers.

One important advantage of the SUV method is that it is very easy to calculate. The required information is the weight of the patient, the administered dose of FDG, the elapsed time from injection to midpoint of the PET image, and the activity in a tissue of interest determined from the PET image. Importantly, the method does not require blood sampling.

Simplified kinetic analysis (8) is a slightly more complex method of quantitation. This approach requires a single blood sample that is used to scale an average blood time activity curve (TAC) so it passes through the value of the single blood sample at the appropriate time. The measure of uptake is then defined as tissue activity divided by the integral of blood activity up to the midpoint of the PET image. This approach is superior to the SUV approach since it relates tissue uptake to the availability of tracer to the tissue via the bloodstream. The SUV approach implicitly assumes the injected activity is uniformly diluted in all patients in the same way, which is probably not true.

Patlak analysis is a more complex analysis scheme (17). It requires dynamic tissue and blood activity data. It is simpler than the most sophisticated analysis method, parameter optimization using the four-parameter Sokoloff/Phelps model (18), but it is generally regarded as too complex for routine clinical use.

In the work presented here we compare these different methods. The patients studied all had colon cancer metastatic to the liver. The clinical questions being addressed include diagnosis of extrahepatic disease and evaluation of response to therapy. The gold standard used was parameter optimization using the four-parameter model with blood TAC data determined from the abdominal aorta scaled using blood activity in late venous blood samples. We compared the simpler measures of FDG metabolism [i.e., SUV with various corrections, simplified kinetic analysis (SKA), and Patlak analysis against FDG metabolic rate (MRFDG) determined with parameter optimization].

The goal of this study was to rank the various data analysis strategies. We have chosen to study a relatively homogeneous group of patients to avoid the confounding problems of different tumors types and different normal tissues. While this study is not likely to be useful in selecting specific threshold values for the uptake parameters, it is quite likely that the rank order of the studies in terms of accuracy will be valid for a wide variety of malignancies.

Section snippets

Patient information

Between July 1997 and December 1998, 40 patients with diagnosed colon cancer metastatic to the liver were studied in the University of Washington PET center using a full-dynamic imaging protocol with venous blood sampling. Twenty-six of the patients were men, and 14 were women. The average age was 62 years (range 27–78). The average weight was 74 kg (range 45–107). The mean injected FDG dose was 9.3 mCi (range 6.9–10.6), and glucose levels ranged from 68 to 263 mg/dL (mean 103).

Six additional

Results

Figure 1 shows the correlation between tumor glucose metabolic rate and Patlak slopes calculated with individual blood TACs and with population-scaled blood TACs, in which the average curve was scaled to match the individual blood activity at 60 min after injection. The correlation is excellent for the Patlak slope plot. Note that the slopes are considerably less than 1.0.

The correlation plots of various measures of uptake plotted against FDGMR determined with parameter optimization (FIG. 2,

Discussion

Quantitation of FDG uptake is essential in oncologic PET studies. FDG metabolic rate has been used as a prognostic indicator that is independent of other measures of tumor grade (9). By setting a quantitative threshold, it is possible to objectively decide if a lesion is benign or malignant (15). Quantitative measurement of FDG uptake has been shown to be useful in a number of malignancies in assessing tumor response to radiotherapy or chemotherapy (12).

Although it is possible to accurately

Acknowledgements

The authors thank Kenneth Krohn and Jeanne Link for radiochemistry support, Thomas Lewellen for physics support, Mark Muzi for assistance with data analysis, and Barbara Lewellen for technical assistance with PET imaging. This work was supported by NIH Grant CA74959.

References (23)

  • C.K Kim et al.

    Standardized uptake values of FDGBody surface area correction is preferable to body weight correction

    J. Nucl. Med.

    (1994)
  • Cited by (142)

    • A tri-modal tissue-equivalent anthropomorphic phantom for PET, CT and multi-parametric MRI radiomics

      2022, Physica Medica
      Citation Excerpt :

      Since DWI are not inherently co-registered with anatomical T2w and T1w images, the non-rigid registration method implemented in the Elastix module of 3D Slicer was used to co-register images [15]. PET images were parameterized in Body-Weighted Standardized Uptake Value [16]. 3D volumes of interest were drawn on each image modality using different segmentation procedures, optimized for the different image characteristics.

    • Assessing Response to Therapy

      2022, Oncologic Imaging: A Multidisciplinary Approach
    • Total-Body PET Kinetic Modeling and Potential Opportunities Using Deep Learning

      2021, PET Clinics
      Citation Excerpt :

      Metabolite correction is needed for some radiotracers36 and can be done by arterial blood sampling (or if the metabolism is slow and the arterial-venous difference is small, venous blood sampling may be sufficient) followed by high-performance liquid chromatography analysis, which is complex and invasive, and therefore is often mainly used to derive a population-based model. Mathematical metabolite correction44–46 is an alternative solution that jointly estimates the parent plasma input function during TAC fitting.47 However, this method requires a complex optimization because of the high nonlinearity of the model and can easily get stuck at a local minimum and has a high computational cost.

    • Quantification of Radiotracer Uptake Into Tissue

      2021, Molecular Imaging: Principles and Practice
    • Analysis of Four-Dimensional Data for Total Body PET Imaging

      2021, PET Clinics
      Citation Excerpt :

      Additionally, a novel approach to embedding the graphical analysis in the reconstruction process has shown good preliminary results from Li and colleagues using a nested Patlak fit within iterations of a DIRECT reconstruction23,24 (Fig. 5). Such images may be especially useful for detection of small or low uptake lesions, especially in the liver where high parenchymal uptake result in difficulty detecting lesions; however, images show that the liver has a low Ki, so liver lesions, which have been shown to have a Ki of 0.08 ± 0.04,25,26 should be much easier to detect on parametric flux images. The higher contrast of lesions on parametric KFDG images from the uEXPLORER has been shown by Wang and colleagues27 in chest, kidney, and thigh lesions.

    • Human brown adipose tissue: Underestimated target in metabolic disease?

      2019, Biochimica et Biophysica Acta - Molecular and Cell Biology of Lipids
      Citation Excerpt :

      To measure tissue substrate uptake rate, a dynamic scan method can be used. A dynamic PET/CT, using Patlak curve [102,103] fitting reveals a measure of glucose uptake rate [35]. For comparison between studies the followed procedures of the combined PET/CT (or PET/MR) requires clear descriptions and preferably standardization.

    View all citing articles on Scopus
    View full text