HAI Book 2025 - Flipbook - Page 121
Bilgel, Murat
10
Positron emission tomography analysis with the dynamicpet Python
package
Murat Bilgel1, Sabri Amer2, Jonghyun Bae3, Mustapha Bouhrara3, Zhaoyuan Gong3, Alex Guo3,
Johnny Uriarte-Lopez2, Kavita Singh1, Granville J. Matheson4,5,6,7, Martin Nørgaard8,9
1
Laboratory of Behavioral Neuroscience, National Institute on Aging, Baltimore, MD, US
Center for Alzheimer’s and Related Dementias, National Institute on Aging, Bethesda, MD, US
3
Magnetic Resonance Physics of Aging and Dementia, Laboratory of Clinical Investigation, National Institute on
Aging, Baltimore, MD, US
4
Department of Biostatistics, Columbia University Mailman School of Public Health, New York, NY, US
5
Department of Psychiatry, Columbia University, New York, NY, US
6
Molecular Imaging and Neuropathology Division, New York State Psychiatric Institute, New York, NY, US
7
Department of Clinical Neuroscience, Center for Psychiatry Research, Karolinska Institutet and Stockholm County
Council, Stockholm, SE
8
Department of Computer Science, University of Copenhagen, Copenhagen, DK
9
Molecular Imaging Branch, National Institute of Mental Health, Bethesda, MD, US
2
The Positron Emission Tomography-Brain Imaging Data Structure (PET-BIDS) (Nørgaard et al. 2022) provides a
standardized file naming and directory organization structure for PET datasets, facilitating scientific
reproducibility and data sharing. It is being extended via the BIDS Extension Proposal 23: PET Preprocessing
Derivatives to establish conventions for image analysis outputs. There is a need for user-friendly PET analysis
tools that can integrate with PET-BIDS datasets. We present dynamicpet, a free and open-source Python package
that provides functions for 4-D PET analysis, including extraction of regional time activity curves (TACs),
denoising, kinetic modeling, and concatenation of interrupted acquisitions.
We demonstrate the functionality of dynamicpet on a 11C-Pittsburgh compound B (PiB) amyloid PET scan acquired
over 70 mins from the Baltimore Longitudinal Study of Aging. Figure 1 illustrates the Python API for loading in a 4D image and calculating its frame duration-weighted temporal mean. The two denoising algorithms provided in
dynamicpet, HYPR-LR (Christian et al. 2010) and NESMA (Bouhrara et al. 2018), are illustrated in Figure 2.
dynamicpet implements several reference region-based models for estimating biological parameters of interest:
standardized uptake value ratio (SUVR), the Logan reference tissue model (LRTM) (Logan et al. 1996), and several
simplified reference tissue model (SRTM) variants (Figure 3). dynamicpet also includes utility functions such as
concatenating interrupted acquisitions by accounting for decay correction differences. Code is available at
https://github.com/bilgelm/dynamicpet.
dynamicpet adds to the growing list of PET analysis software developed to work with PET-BIDS data, such as
PET2BIDS (Galassi et al. 2024), petdeface (https://github.com/openneuropet/petdeface), kinfitr (Matheson 2019),
bloodstream (https://github.com/mathesong/bloodstream), and petprep
(https://github.com/mnoergaard/petprep_hmc). Its simple framework facilitates the implementation of
additional kinetic models. Future work will expand the collection of kinetic models and implement quality control
checks for the inputs and outputs of kinetic models.
HAI2025 - 121
