Automated PET tracer classification in multi-site, multi-cohort studies using deep learning
Year: 2026
Type: article
Abstract: BACKGROUND: Radiotracer labels in positron emission tomography (PET) DICOM metadata are manually entered and often unreliable or missing, requiring expert visual verification for downstream analyses. We developed and validated deep learning models to classify PET tracers directly from image data across multiple sites and cohorts. RESULTS: We used PET scans obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 7429) and the Standardized Centralized Alzheimer's & Related Dementias Neuroimaging study (SCAN, n = 4869) to classify 10 tracers. We evaluated generalizability on an external test set of Mayo Clinic Study of Aging and ADRC scans containing four tracers (n = 10558). ADNI and SCAN data were combined and split into training (60%), validation (20%), and internal test (20%) sets. We trained two ConvNeXt-Large models initialized with ImageNet weights: a 10-class model (AV1451, FBP, FBP-Early, FBB, FBB-Early, FDG, MK6240, NAV4694, PI2620, PIB) and a 3-class model grouping tracers by biological target (Amyloid, FDG, Tau). Performance was evaluated using macro-F1, average precision (AP), and confusion matrices, Integrated Gradients for attribution, with Uniform Manifold Approximation and Projection (UMAP) and cosine similarity to assess model interpretability and multi-cohort effects on the embedding structure. The 3-class model achieved macro-F1 ≥ 0.99 across all test sets. The 10-class model achieved macro-F1 of 0.91 on the internal test and 0.93 on the external test set where only four tracers were available. Lower performance was observed for NAV4694 (AP 0.70; misclassified as FBB, 17.2%), PI2620 (0.79; AV1451, 20.0%), and FBB (0.80; FBP, 23.3%). The attribution maps indicated the models learned some tracer specific uptake patterns as well as general imaging features. Class embeddings in the UMAP projection showed clear separation, with limited overlap among underrepresented tracers with similar binding characteristics. Cohort-level differences in learned representations were most pronounced among tau tracers, with higher within-class cosine similarity in ADNI than in SCAN (mean [SD], 0.26 [0.10] vs. 0.01 [0.06]), while amyloid tracers showed minimal shift across cohorts. CONCLUSIONS: Classification was near-perfect when grouping tracers into three biological targets, and the 10-class model generalized well to the four tracers available in the independent external cohort. A few tracers with similar binding characteristics showed some overlap in latent space, with minimal impact on overall classification performance. Our results demonstrate that deep learning models can identify PET tracers from image appearance alone and have the potential to support semi-automated tracer identification and data curation in large multi-site PET studies.
Source: EJNMMI Research
Institution: Mayo Clinic
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $3,190
Funders: Emory University, Mayo Foundation for Medical Education and Research, GHR Foundation, Elsie and Marvin Dekelboum Family Foundation, National Institutes of Health, National Institute on Aging, National Institute of Mental Health, National Institute on Deafness and Other Communication Disorders, National Institute of Neurological Disorders and Stroke, National Institute of Biomedical Imaging and Bioengineering, National Center for Research Resources
Awards: Multimorbidity and Aging: Rochester Epidemiology Project, Role of age, dopamine, and tau related network disruption in setting a context for progression toward Alzheimer's disease, Michigan Alzheimer's Disease Core Center, UC Davis Alzheimer's Disease Core Center, Modeling Axonal Density and Inflammation-Associated Cellularity in Alzheimer’s Disease Using Hybrid Diffusion Imaging, Indiana Alzheimer Disease Center, Alzheimer's Disease Research Center, G1711016037, G1835837428, Mapping Abnormal Neurodevelopment in Schizophrenia, Functional-anatomic exploration of cognitive control, Tau, amyloid, & white matter burden interact to impact brain networks in preclinical Alzheimer's disease, Determinants of Neurodegenerative Decline in Primary Progressive Aphasia, Identification of cognitive decline and dementia: Prediction by everyday driving behaviors and physiological responses, Multi-Domain Sensory Measures as Biomarkers of Alzheimer's Disease in Preclinical and Prodromal Stages, Early Onset AD Consortium - the LEAD Study (LEADS), Morphometry Biomedical Informatics Research Network, Alzheimer's Disease Research Center, Use of the KF Modified Story Memory Technique to Improve New Learning and Memory in Individuals with Mild Cognitive Impairment, Slow wave sleep (SWS) and the effect of African ancestry on amyloid burden, a longitudinal study, 1Florida Alzheimer's Disease Research Center, G3486421421, G3638781909, Alzheimers Disease Patient Registry, G3930010334, DETECT-AD: Digital Evaluations and Technologies Enabling Clinical Translation for Alzheimer's Disease, Nevada Exploratory Alzheimer's Disease Research Center (NVEADRC), Cleveland Alzheimer's Disease Research Center, The Alzheimer's Disease Research Center at the University of California, Irvine, Stanford Alzheimer's Disease Research Center, Rush Alzheimer's Disease Core Center, DRIVING PERFORMANCE IN PRECLINICAL ALZHEIMER'S DISEASE, G5484535055, Language in Primary Progressive Aphasia, G5900136256, Upgrade for MRI Instrument for Functional Brain Imaging, PET Imaging of Synaptic Density in Alzheimers Disease, Alzheimer's Disease Research Center - Overall, Wisconsin Alzheimer's Disease Research Center, University of Kansas Alzheimer's Disease Core Centers, New Mexico Alzheimer's Disease Research Center, Alzheimer Disease Research Center, Yale Alzheimer Disease Research Center, Arizona Alzheimer's Disease Core Center, Healthy Aging and Senile Dementia, Washington University Institute of Clinical and Translational Sciences, EXTRAMURAL RESEARCH FACILITIES IMPROVEMENT PROGRAM, Memory Circuitry in MCI and Early Alzheimerâs Disease Prodrome: Molecular Drivers, PET Imaging of SV2A and Other Biomarkers in Alzheimer's Disease, Boston University Alzheimer's Disease Center, Alzheimer's Disease Research Center, Vanderbilt Alzheimer's Disease Research Center, Alzheimer's Disease Core Center, Sleep Aging and Risk for Alzheimer's disease-Resubmission-1, NINDS CENTER CORE FOR BRAIN IMAGING, Protecting the Confidentiality of Participants in Studies of Alzheimer's Disease and Related Dementias by Replacing Face Imagery in MRI, University of Florida - Mt. Sinai Medical Center AD Research Center, Determinants of neurodegenerative decline in the aphasic variant of Alzheimer's disease, Modulators of Medial Temporal Lobe Subregion Structure and Function in Normal and Pathological Aging, Exceptional Cognitive Aging: Neuropsychologic, Anatomic and Pathologic Correlates, MarkVCID Validation in the General Community, G8611190880, Treating mild cognitive impairment with transcranial direct current stimulation, G9000150898, G9000338148, ARTFL LEFFTDS Longitudinal Frontotemporal Lobar Degeneration, Cycle 2 (ALLFTD2), Disease pathways in the population determined by amyloid, tau, and neurodegeneration imaging biomarkers, UCSD Shiley-Marcos Alzheimer's Disease Research Center P30, Michigan Alzheimer's Disease Research Center, Investigating Resistance and Resilience Mechanisms in Alzheimer’s Disease, Oregon Alzheimer's Disease Research Center, Alzheimer's Disease Research Center, UC Davis Alzheimer's Disease Research Center, Johns Hopkins Alzheimer's Disease Research Center, Indiana Alzheimer's Disease Research Center, New Approaches to Dementia Heterogeneity, ASYMMETRIC NEURODEGENERATION AND LANGUAGE IN PRIMARY PROGRESSIVE APHASIA, Arizona Alzheimer's Disease Research Center, Stanford Alzheimer's Disease Research Center, Study to Uncover Pathways to Exceptional Cognitive Resilience in Aging (SUPERAging), APOE in the Predisposition to, Protection from and Prevention of Alzheimer's Disease, Mount Sinai Alzheimer's Disease Research Center, Cognitive SuperAging: A model to explore resilience and resistance to aging and Alzheimers disease, Longitudinal Imaging Biomarkers of Prodromal DLB, Duke/UNC Alzheimer's Disease Research Center, THE XNAT IMAGING INFORMATICS PLATFORM, SCAN: Standardized Centralized Alzheimer's and Related Dementias Neuroimaging, National Alzheimer's Coordinating Center, Antecedent Biomarkers for Alzheimer Disease: The Adult Children Study, Alzheimer's Disease Research Center, Research Program on Cognition and Neuromodulation Based Interventions (RP-CNBI), Goizueta Alzheimer's Disease Research Center, The Clinical Significance of Incidental White Matter Lesions on MRI Amongst a Diverse Population with Cognitive Complaints (INDEED), Mayo Alzheimer's Disease Research Center
OpenAlex ID: W7171501991