HAI Book 2025 - Flipbook - Page 573
Ruppert, Emma
131
Harmonizing visual readings of Flortaucipir and MK-6240 using headto-head data
Emma Ruppert1, Marina Medeiros1, Andreia Rocha1, Carolina Soares1, Livia Amaral1, Guilherme
Povala1, Guilherme Bauer-Negrini1, Pamela L. Ferreira1, Firoza Lussier1, Markley Oliveira Jr.1,
Matheus Scarpatto Rodrigues1, Rayan Mroue1, Hussein Zalzale1, Douglas Leffa1, Pampa Saha1,
Bruna Bellaver1, Dana L. Tudorascu1, David Soleimani-Meigooni4, Juan Fortea5, Val Lowe6,
Hwamee Oh7, Belen Pascual2, Brian A. Gordon8, Pedro Rosa-Neto9, Suzanne Baker10, Tharick
A. Pascoal1
1
University of Pittsburgh, Department of Psychiatry, Pittsburgh, PA, US
Houston Methodist Research Institute, Department of Neurology, Houston, TX, US
3
University of California San Francisco, Memory and Aging Center, San Francisco, CA, US
4
Hospital de la Santa Creu i Sant Pau, Sant Pau Memory Unit, Department of Neurology, Barcelona, ES
5
Mayo Clinic, Department of Radiology, Rochester, MN, US
6
Brown University, Department of Psychiatry and Human Behavior, Providence, RI, US
7
Washington University in St. Louis, Department of Radiology, St. Louis, MO, US
8
Translational Neuroimaging Laboratory, McGill University Research Centre for Studies in Aging, Douglas Research
Institute, Montréal, QC, CA
9
Lawrence Berkeley National Laboratory, Berkeley, CA, US
2
Background: Various visual reading methods for detecting brain tau-PET deposition have been developed for
different tracers. However, their comparative performance in the same individuals has not been explored. This
study aims to compare the performance of existing visual reading methods using [18F]Flortaucipir(FTP) and
[18F]MK-6240(MK) and provide preliminary data toward a unified visual reading approach for all tau-PET tracers.
Methods: We assessed 340 individuals across the AD spectrum from the HEAD cohort (CU-n=175; CI-n=143). A
subset of 101 individuals (CU-n=52; CI-n=49) was used to develop a unified visual reading approach. All participants
underwent MRI and PET scans with FTP and MK. Visual reads, based on previous methods, were independently
performed by two experts blinded to participants' demographics and diagnoses. For FTP, we applied Sonni's
(PMID-33313377) method, and for MK, we used Shuping9s(PMID-36873926) method. Inter-rater agreement was
evaluated using unweighted Cohen9s kappa(κ). Differences between visual reading groups were assessed using
ANOVA with post-hoc Tukey tests.
Results: Using previous methods, inter-rater agreement was 72.1% for FTP (κ=0.55[0.48-0.62]) and 83.5% for MK
(κ=0.69[0.62-0.76]). Agreement between the two tracers' visual reads(Fig.1) ranged from 65.6%-75.3%. These
methods classified tau burden as negative, non-AD-like, low or high. In contrast, the developed method achieved
an inter-tracer agreement of 87.1% and classified tau burden as negative, low, moderate, or high(Fig.2).
Additionally, mean differences in tau-PET SUVR in the medial temporal lobe(MTL), plasma p-tau217, and MoCA
scores across visual classification groups demonstrated greater statistical significance when using the
developed method(Fig.3).
Conclusion: This study demonstrates that current visual reading methods produce varying classifications
depending on the tracer used, revealing discrepancies below desired clinical performance standards. Our method
improved concordance between visual reads by developing a unified, tracer-agnostic approach. Importantly, this
method has the potential to create a clinician-friendly single visual reading technique that could be widely
adopted for all tau tracers.
HAI2025 - 573
