Quantifying extracerebral off‐target contamination in [ 18 F]MK6240 PET imaging using compartmental modeling

Jessie Fanglu Fu, Cristina Lois, Amal Tiss, Hasan Sari, Alex Becker, Justin S. Sanchez, Nicolas J. Guehl, Marc D. Normandin, Georges El Fakhri, K. N. Johnson, Julie C. PriceAlzheimer 2025 · DOI

TL;DR

This paper introduces a new compartmental modeling technique to accurately quantify and correct for "off-target" signals in PET scans used to detect Alzheimer's disease (AD) tau pathology. By accounting for signal contamination from outside the brain, this method promises to improve the sensitivity and accuracy of early AD diagnosis.

What Problem Does This Paper Address?

Detecting early signs of Alzheimer's disease (AD) is crucial for timely intervention. Positron Emission Tomography (PET) imaging with tracers like [<sup>18</sup>F]MK6240 shows promise for visualizing tau tangles, a hallmark of AD. However, these tracers can also bind to tau in areas outside the brain, such as the meninges (the membranes surrounding the brain), leading to "extracerebral off-target contamination." This contamination can obscure or mimic the signal from actual tau pathology within the brain, potentially leading to misdiagnosis or reduced sensitivity in detecting early-stage AD. Existing methods like partial volume correction (PVC) help, but they may not fully address this specific type of contamination.

Methods

The researchers developed and evaluated an enhanced compartmental modeling approach. They started with a standard two-tissue compartmental model (2TCM) and introduced an alternative model, termed 2TCM+s. The key innovation in 2TCM+s is the addition of an "extracerebral term" (sC<sub>extracerebral</sub>) that specifically accounts for the signal originating from outside the brain. This term quantifies the fractional contribution of this extracerebral signal. They compared the performance of 2TCM+s against the standard 2TCM and a kernel-based PVC method. Model performance was assessed using metrics like the Akaike Information Criteria (AIC), distribution volume (V<sub>T</sub>), distribution volume ratio (DVR), and standardized uptake value ratios (SUVR) in late frames.

Datasets

The study utilized PET imaging data from 13 participants: 8 cognitively normal (CN) individuals and 5 individuals with mild cognitive impairment (MCI). These participants underwent 2-hour [<sup>18</sup>F]MK6240 PET scans, along with MRI and arterial sampling for kinetic modeling. The study did not introduce a new benchmark dataset but rather evaluated their method on existing clinical data.

Key Insights and Ideas

  • Extracerebral signal is a real issue: The study visually and quantitatively confirmed that [<sup>18</sup>F]MK6240 PET scans do exhibit significant signal from extracerebral sources, particularly in regions close to the meninges.
  • 2TCM+s is a powerful correction tool: The novel 2TCM+s model proved to be the most effective at accounting for this extracerebral contamination, outperforming both the standard 2TCM and kernel-based PVC in terms of model fit (AIC).
  • Correcting the signal matters for interpretation: By accounting for extracerebral contamination, the 2TCM+s model led to lower estimates of V<sub>T</sub>, especially in highly contaminated regions. This suggests that previous estimates might have been inflated by off-target binding.
  • Improved diagnostic sensitivity: The corrected SUVR values from the 2TCM+s model showed a larger effect size for differentiating between cognitively normal individuals and those with MCI, indicating a potential improvement in diagnostic sensitivity for early AD detection.

Key Results

  • The 2TCM+s model demonstrated the best AIC, indicating a superior fit to the PET data compared to 2TCM and kernel-based PVC.
  • The estimated extracerebral contamination term (s) ranged from 0.16 to 0.43, with higher values observed in regions like the entorhinal cortex, which is near the meninges.
  • Accounting for extracerebral contamination with 2TCM+s resulted in lower V<sub>T</sub> values, particularly in highly contaminated areas.
  • For differentiating CNs from MCIs, the SUVR90-110min corrected with 2TCM+s showed the largest effect size (Cohen's d = -1.2), surpassing kernel-PVC (-1.0) and non-PVC (-0.58) SUVR.

Limitations and Future Work

The authors acknowledge that future studies should include a larger number of participants to further validate these findings. Additionally, they plan to extend this compartmental modeling approach to [<sup>18</sup>F]MK6240 PET data acquired without dynamic acquisition (i.e., using only late-frame summation images). This would broaden the applicability of their correction method to more routine clinical and research settings where dynamic scans are not always feasible.

Summary updated

Abstract

A compartmental modeling approach quantifies regional [<sup>18</sup>F]MK6240 extracerebral contamination, potentially improving V<sub>T</sub>, DVR, and SUVR accuracy and sensitivity for early tau detection, even compared to PVC data. Future studies will include more participants and extend these corrections to [<sup>18</sup>F]MK6240 data without dynamic acquisition, broadening applicability in clinical and research settings.
BibTeX
@inproceedings{fu2025quantifying,
  author = {Jessie Fanglu Fu and Cristina Lois and Amal Tiss and Hasan Sari and Alex Becker and Justin S. Sanchez and Nicolas J. Guehl and Marc D. Normandin and Georges El Fakhri and K. N. Johnson and Julie C. Price},
  title = {{Quantifying extracerebral off‐target contamination in [ <sup>18</sup> F]MK6240 PET imaging using compartmental modeling}},
  booktitle = {Alzheimer},
  year = {2025},
  doi = {10.1002/alz70856\_099911},
}
Figure