Automated PET tracer classification in multi-site, multi-cohort studies using deep learning.
5 affiliations
- 1Aging and Dementia Imaging Research (ADIR), Department of Radiology, Mayo Clinic, 200 First Street SW, Rochester, MN, 55905, USA. [email protected].
- 2Department of Information Technology, Mayo Clinic, Rochester, MN, 55905, USA.
- 3Aging and Dementia Imaging Research (ADIR), Department of Radiology, Mayo Clinic, 200 First Street SW, Rochester, MN, 55905, USA.
- 4Department of Nuclear Medicine, Mayo Clinic, Rochester, MN, 55905, USA.
- 5Department of Neurology, Mayo Clinic, Rochester, MN, 55905, USA.
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.
Related
- Curriculum-Based Deep Learning for Automated Segmental Left Ventricular Hypertrophy Classification Using Short-Axis Echocardiography
- Automated brain MRI image classification through machine learning and deep learning approaches
- A trustworthy anatomically-guided region-aware deep learning framework for automated Alzheimer’s disease classification from brain MRI
- Automated acquisition of thyroid cytology image features using self-supervised learning and application to image classification
- Code-Free Automated Machine-Learning (AutoML) Classification of Curated MRI Images Labeled for Rare Neurological Diseases: An Image-Level Proof-of-Concept Study
BibTeX
@article{gebre2026automated,
title = {{Automated PET tracer classification in multi-site, multi-cohort studies using deep learning.}},
author = {Robel K Gebre and Robert I Reid and Matthew L Senjem and Jeffrey L Gunter and Val J Lowe and Kejal Kantarci and Ronald C Petersen and Jonathan Graff-Radford and Clifford R Jack and Prashanthi Vemuri and Christopher G Schwarz and Alzheimer’s Disease Neuroimaging Initiative},
journal = {EJNMMI research},
year = {2026},
month = jul,
doi = {10.1186/s13550-026-01486-w},
url = {https://europepmc.org/article/MED/42909731},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Automated PET tracer classification in multi-site, multi-cohort studies using deep learning.
AU - Robel K Gebre
AU - Robert I Reid
AU - Matthew L Senjem
AU - Jeffrey L Gunter
AU - Val J Lowe
AU - Kejal Kantarci
AU - Ronald C Petersen
AU - Jonathan Graff-Radford
AU - Clifford R Jack
AU - Prashanthi Vemuri
AU - Christopher G Schwarz
AU - Alzheimer’s Disease Neuroimaging Initiative
PY - 2026
DA - 2026/07/01
JO - EJNMMI research
DO - 10.1186/s13550-026-01486-w
UR - https://europepmc.org/article/MED/42909731
AB - 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.
SN - 2191-219X
ER -