Correcting SUVR bias by accounting for radiotracer clearance in tissue: A validation study with [ 18 F]FE-PE2I PET in cross-sectional, test-retest and longitudinal cohorts

Minyoung Oh, Praveen Honhar, Richard E. Carson, Ansel T. Hillmer, Andrea VarroneJournal of Cerebral Blood Flow & Metabolism 2025 · DOI

TL;DR

This study introduces and validates a new method, SUVRc, to correct for biases in Positron Emission Tomography (PET) imaging of dopamine transporters (DAT) in Parkinson's disease. By accounting for how the radiotracer clears from tissues, SUVRc significantly reduces overestimation of DAT levels compared to standard methods, offering a more accurate measure for both cross-sectional and longitudinal studies.

What Problem Does This Paper Address?

Parkinson's disease (PD) is characterized by the degeneration of dopaminergic neurons, leading to a reduction in dopamine transporters (DAT). PET imaging, particularly with the radiotracer [<sup>18</sup>F]FE-PE2I, is a crucial tool for quantifying DAT levels in vivo and tracking disease progression. A common quantification method is the Standardized Uptake Value Ratio (SUVR), which compares tracer uptake in a target region (like the striatum) to a reference region (like the cerebellum). However, standard SUVR measurements, especially when taken over longer time windows (e.g., 50-80 minutes post-injection), are known to be biased. This bias arises because the radiotracer doesn't reach a true equilibrium in the tissues within this timeframe; it continues to clear, leading to an overestimation of DAT binding. This overestimation can obscure subtle changes, particularly in longitudinal studies tracking disease progression or treatment response. The paper aims to validate a novel correction method, SUVRc, designed to mitigate this bias by explicitly modeling and correcting for radiotracer clearance.

Methods

The core of this study is the validation of a new correction method, SUVRc, for Standardized Uptake Value Ratio (SUVR) measurements in [<sup>18</sup>F]FE-PE2I PET imaging. The SUVRc formula is derived to account for the clearance rates of the radiotracer from both the target (β_tar) and reference (β_ref) tissues.

The SUVRc calculation relies on:

  • SUVR: The standard ratio of tracer uptake in the target tissue to the reference tissue over a specific time window.
  • Clearance Rates (β_tar, β_ref): These represent how quickly the radiotracer leaves the target and reference tissues, respectively. β_ref is estimated from a single exponential fit to the reference tissue's time-activity curve. β_tar is estimated using a regression-based approach that relates it to the clearance rate of a composite region (a larger, high-uptake area like the caudate and putamen) and the difference in SUVR between the composite and target regions. This approach aims to overcome the noise sensitivity of directly estimating β_tar from small regions.
  • Kinetic Parameters (k2,ref, R1): These are assumed based on prior literature.

The study then rigorously validated SUVRc against established measures like Distribution Volume Ratio (DVR) – considered the gold standard – and early SUVR (15-45 min) across three distinct cohorts:

  1. Cross-sectional: Comparing PD patients and healthy controls (HC).
  2. Test-retest: Assessing the reliability of measurements over short intervals.
  3. Longitudinal: Evaluating changes over a longer period (up to 2 years).

Key metrics used for comparison included:

  • Bias: The percentage difference between the measured value and DVR.
  • Discriminative Power: Assessed using Cohen's d, indicating how well a measure differentiates PD patients from HC.
  • Test-Retest Variability (TRV): The consistency of measurements over repeated scans.
  • Annual Longitudinal Change (ALC): The rate of change in DAT binding per year.
  • Receiver Operating Characteristic (ROC) analysis: To evaluate diagnostic performance (Area Under the Curve - AUC).

Datasets

The study utilized data from three distinct cohorts of participants:

  • Cross-sectional Cohort: 38 Parkinson's Disease (PD) patients and 38 age- and sex-matched healthy controls (HC).
  • Test-Retest Cohort: 9 PD patients who underwent repeat PET scans.
  • Longitudinal Cohort: 21 PD patients who underwent PET scans at baseline and a follow-up approximately 2 years later.

These datasets were previously collected and partially reported in earlier studies, but were re-analyzed using the novel SUVRc method. No new benchmark datasets were introduced.

Key Insights and Ideas

  • The "Late Window" SUVR Problem: The study strongly reinforces that SUVR measurements taken in the later time windows (50-80 min) are significantly biased, systematically overestimating DAT binding. This is a critical insight for anyone using PET imaging for DAT quantification.
  • SUVRc as a Powerful Bias Corrector: The proposed SUVRc method effectively corrects this bias, bringing the values much closer to the gold standard (DVR) in both healthy controls and PD patients. This is achieved by explicitly modeling the radiotracer's clearance from tissues.
  • Improved Accuracy, Not Necessarily Precision: While SUVRc significantly improves the accuracy of DAT quantification by reducing bias, it can sometimes exhibit slightly higher test-retest variability compared to DVR or early SUVR. This is attributed to the sensitivity of the underlying β_tar estimation to small fluctuations.
  • Longitudinal Potential: Despite the slightly higher TRV, SUVRc shows comparable annual longitudinal changes to DVR, suggesting it holds promise for tracking disease progression or treatment effects over time, provided the β_tar estimation is robust.
  • β_tar as a Key Indicator: The study highlights β_tar (clearance rate from the target tissue) as a crucial parameter. Its estimation, while challenging, is central to the SUVRc correction. The variability in β_tar across regions and individuals is a key factor influencing the performance of SUVRc.

Key Results

  • Bias Reduction: Standard SUVR (50-80 min) showed substantial positive bias in both HC (mean 48.31%) and PD (mean 29.91%). SUVRc dramatically reduced this bias to near zero in both groups (HC: -0.80%, PD: -0.12%), with statistically significant improvements (p < 0.001).
  • Discriminative Power: Cohen's d values, indicating the ability to differentiate PD from HC, were similar across DVR, SUVRc, and early SUVR (15-45 min) in the striatum, suggesting SUVRc maintains good diagnostic power.
  • Test-Retest Variability (TRV): DVR (3.2%) and early SUVR (3.9%) showed lower TRV than SUVRc (6.8%) and standard late SUVR (6.4%). The higher TRV for SUVRc was linked to the variability in β_tar estimation.
  • Longitudinal Changes (ALC): The annual longitudinal percent change for SUVRc (4.4%) was comparable to DVR (4.5%) and early SUVR (4.2%), with a lower standard deviation than late SUVR (5.2%).
  • Correlation: SUVRc showed strong correlations with DVR across regions (r² up to 0.93), similar to early SUVR.

Limitations and Future Work

  • β_tar Estimation Sensitivity: The primary limitation identified is the sensitivity of SUVRc to errors in the estimation of β_tar. While the regression-based method is an improvement, minor changes in β_tar can lead to increased bias in SUVRc, contributing to its higher test-retest variability compared to DVR.
  • Longitudinal Caution: Due to the β_tar sensitivity, the authors advise caution when using SUVRc for longitudinal studies, emphasizing the need for robust β_tar estimation.
  • Future Directions:
    • Improving β_tar Estimation: The authors suggest exploring more sophisticated methods for estimating β_tar, such as incorporating quadratic terms for regions with markedly different kinetics (like the SN), using Bayesian strategies with stronger priors, or employing deep neural networks.
  • Broader Applicability: The study suggests that this SUVRc correction approach could be valuable for other PET tracers used in diagnostic imaging (e.g., amyloid or tau PET) and for tracers with slower kinetics, potentially improving accuracy and consistency across different centers and acquisition protocols.
  • Prognostic Marker Potential: The method could be important for accurate quantification in prognostic markers, especially for drugs that show changes in PET binding over time.

Summary updated

Abstract

Quantification of dopamine transporter (DAT) with [<sup>18</sup>F]FE-PE2I PET is an important progression marker for Parkinson's disease (PD). This study aimed to validate a novel correction (<i>SUVR</i>c) for a less-biased estimate of <i>SUVR</i> by accounting for [<sup>18</sup>F]FE-PE2I clearance-rate, in independent cross-sectional (38 PD, 38 controls), test-retest (9 PD) and longitudinal cohorts (21 PD). <i>SUVR</i>c was calculated as <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mfrac><mml:mrow><mml:mi>SUVR</mml:mi></mml:mrow><mml:mrow><mml:mn>1</mml:mn><mml:mi> </mml:mi><mml:mo>-</mml:mo><mml:mi> </mml:mi><mml:mfrac><mml:mrow><mml:msub><mml:mrow><mml:mi>β</mml:mi></mml:mrow><mml:mrow><mml:mi>ref</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mrow><mml:msub><mml:mrow><mml:mi>k</mml:mi></mml:mrow><mml:mrow><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mi>ref</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:mfrac><mml:mi> </mml:mi><mml:mo>+</mml:mo><mml:msub><mml:mrow><mml:mi>β</mml:mi></mml:mrow><mml:mrow><mml:mi>tar</mml:mi></mml:mrow></mml:msub><mml:mfrac><mml:mrow><mml:mi>SUVR</mml:mi></mml:mrow><mml:mrow><mml:msub><mml:mrow><mml:mi>k</mml:mi></mml:mrow><mml:mrow><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mi>ref</mml:mi></mml:mrow></mml:msub><mml:msub><mml:mrow><mml:mi>R</mml:mi></mml:mrow><mml:mrow><mml:mn>1</mml:mn></mml:mrow></mml:msub></mml:mrow></mml:mfrac></mml:mrow></mml:mfrac></mml:math>. <i>β</i><sub>tar</sub> and <i>β</i><sub>ref</sub> are the clearance rates from the target and reference tissues. Bias relative to <i>DVR</i>, discriminative power, test-retest variability (TRV) and annual longitudinal change (ALC) were used to compare <i>SUVR</i><sub>50-80 min</sub>, <i>SUVR</i>c<sub>50-80 min</sub>, <i>SUVR</i><sub>15-45 min</sub> and <i>DVR. SUVR</i><sub>50-80 min</sub> showed high bias across all regions (HC: mean: 48.31 ± 20.49% [range: 28.32-53.80%]; PD: 29.91 ± 13.95% [20.45-39.80%]) that was corrected by <i>SUVR</i>c<sub>50-80 min</sub> (HC: -0.80 ± 12.72% [-9.69-11.64%]; PD: -0.13 ± 7.41% [-5.04-2.97%]), <i>p < </i>0.001 for both groups compared to mean bias of <i>SUVR</i><sub>50-80 min</sub>, similar to <i>SUVR</i><sub>15-45 min</sub>. For the striatum, Cohen's <i>d</i> was similar for all measures. TRV were 3.2 ± 2.5% (<i>DVR</i>), 6.4 ± 5.7% (<i>SUVR</i><sub>50-80 min</sub>), 6.8 ± 5.9% (<i>SUVR</i>c<sub>50-80 min</sub>) and 3.9 ± 3.2% (<i>SUVR</i><sub>15-45 min</sub>). Higher TRV of <i>SUVR</i>c<sub>50-80 min</sub> was due to TRV of 9.2 ± 5.1% [1.1-19.4] for β<sub>tar</sub>. ALC was 4.5 ± 4.2% (<i>DVR</i>), 5.2 ± 6.5% (<i>SUVR</i><sub>50-80 min</sub>), 4.4 ± 4.1% (<i>SUVR</i>c<sub>50-80 min</sub>) and 4.2 ± 4.1% (<i>SUVR</i><sub>15-45 min</sub>). <i>SUVR</i>c<sub>50-80 min</sub> reduced bias compared to <i>SUVR</i><sub>50-80 min</sub>, as previously reported. <i>SUVR</i>c<sub>50-80 min</sub> was sensitive to small changes of β<sub>tar</sub>, with higher TRV compared to <i>DVR</i>, but with similar ALC, suggesting that it can reliably assess longitudinal DAT changes.
BibTeX
@article{oh2025correcting,
  author = {Minyoung Oh and Praveen Honhar and Richard E. Carson and Ansel T. Hillmer and Andrea Varrone},
  title = {{Correcting <i>SUVR</i> bias by accounting for radiotracer clearance in tissue: A validation study with [ <sup>18</sup> F]FE-PE2I PET in cross-sectional, test-retest and longitudinal cohorts}},
  journal = {Journal of Cerebral Blood Flow \& Metabolism},
  year = {2025},
  doi = {10.1177/0271678x251322407},
}

Loading related content...

Figure