HAI Book 2025 - Flipbook - Page 358
Minhas, Davneet
78
An accurate and automated technique for removing spill-over
contamination and off-target retention from a cerebellar reference region
for tau PET
Davneet Minhas1, Alex DelBene1, Alexandra Gogola1, Alexandria Reese1, Weiquan Luo1, Charles
Laymon1, Brian Lopresti1, Vincent Dore2, Tharick Pascoal1, Suzanne Baker3, Dana Tudorascu1,
Victor Villemagne1
1
University of Pittsburgh, Pittsburgh, PA, US
Austin Health, Melbourne, AU
3
Lawrence Berkeley National Laboratory, Berkeley, CA, US
2
Background: Reference region (RR) definition impacts tau PET quantification. In this work, we explore methods
of optimizing the RR by minimizing spillover and off-target signal.
Methods: We developed CerTauSeg, an algorithm that identifies clusters of contamination in an individual
subject9s whole cerebellar gray tau PET scan with no a priori knowledge. CerTauSeg accuracy was evaluated
against standard RRs using simulations, which included varying sources of off-target contamination (superior
cerebellum, meninges) and sinogram-added noise, constructed with Siemens E7 Tool Projector software.
CerTauSeg was applied to 338 ADNI FTP-PET scans in template space to generate subject-specific
