Automated joint segmentation enables quantitative molecular imaging of degeneration with [¹⁸F]FDG and [¹⁸F]NaF PET/CT
6 affiliations
- 1Department of Radiology, Hospital of the University of Pennsylvania, Philadelphia, 19104 PA USA
- 2Drexel University College of Medicine, Philadelphia, PA USA
- 3Sidney Kimmel Medical College, Thomas Jefferson University, Philadelphia, PA USA
- 4First Faculty of Medicine, Charles University, Prague, Czechia Czechia
- 5Department of Nuclear Medicine, Odense University Hospital, Odense, Denmark
- 6Department of Clinical Research, University of Southern Denmark, Odense, Denmark
Abstract
Purpose: Manual quantification of joint metabolism and inflammation with PET is time-consuming and difficult to standardize due to ambiguous joint boundaries. We aimed to develop an automatic joint segmentation method that efficiently and reliably quantifies activity in weight-bearing joints using [18F]fluorodeoxyglucose (FDG) and sodium [18F]fluoride (NaF) PET/computed tomography (PET/CT).
Methods: Segmentations were generated from CT scans using the TotalSegmentator model. From these, 30 weight-bearing joints were segmented by identifying the region between adjacent bones within a set distance threshold. Scans of healthy volunteers were analyzed (75 with FDG, 80 with NaF). Mean standardized uptake value (SUVmean) was calculated for each joint region, and Pearson correlation analysis was used to evaluate relationships with age and body mass index (BMI). Benjamini–Hochberg false discovery rate (FDR) correction was applied within each tracer–covariate family.
Results: The automated approach reduced segmentation time from 30 to ~ 1.5 min per scan. Joint ROI volumes generated independently from the two PET/CT sessions of the same participant were highly concordant (median r = 0.98, median intraclass correlation coefficient 0.94). After false discovery rate correction, NaF uptake was positively correlated with BMI in the thoracic (T3-T6, T7-T8, T9-T11) and lumbosacral (L5-S1) intervertebral discs (IVDs) and in the sacroiliac (SI), hip and knee joints, and positively with age in the cervical IVDs (C1-C2, C3-C7), with negative age correlations at T4-T5, T6-T7, L1-L2 and in both SI joints (all q < 0.05). FDG uptake was positively correlated with BMI in the IVDs from C1-T6 (other than T1-T2) and T9-S1, in both SI joints, the left hip and both knees, and with age in both knees, and negatively with age at T2-T3 and in both SI joints (all q < 0.05). Associations meeting p < 0.05 but not q < 0.05 are identified in Tables 2, 3, 4 and 5. The positive cervical NaF-age associations arise in the only joints where SUVmean was appreciably correlated with ROI size and are therefore reported as provisional.
Conclusion: Our segmentation approach rapidly quantifies joint metabolism and revealed significant associations between PET tracer uptake, BMI, and age. It offers a standardized framework for future PET/CT joint analysis.
Clinical trial registration: Cardiovascular Molecular Calcification Assessed by 18 F-NaF PET/CT (CAMONA), ClinicalTrials.gov NCT01724749. https://clinicaltrials.gov/study/NCT01724749.
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1007/s00259-026-08139-z.
Related
- Automated quantitative computed tomography grading of lumbar facet joint degeneration: influence of segmentation errors on instance-level clinical assessment.
- Molecular Fingerprinting Enables Mass Quantification and Semi-Quantitative Source Apportionment of Black Carbon in Biological Samples and Beyond
- Automated radiomics pipeline integrating TotalSpineSeg for Pfirrmann grading of lumbar disc degeneration with SHAP-based interpretability
- Automated thigh muscle segmentation using foundation models: validation and impact of segmentation quality
- CT Ventricular Volume Reference Percentile Curves Using Automated nnU-Net Segmentation: A Study with LLM-Facilitated Cohort Selection
BibTeX
@article{lee2026automated,
title = {{Automated joint segmentation enables quantitative molecular imaging of degeneration with [¹⁸F]FDG and [¹⁸F]NaF PET/CT}},
author = {William K. Lee and Shiv Patil and Darshil H. Patel and Zhuorui Li and Ondrej Fanta and Om H. Gandhi and Thomas J. Werner and Poul F. Høilund-Carlsen and Abass Alavi},
journal = {European Journal of Nuclear Medicine and Molecular Imaging},
year = {2026},
month = aug,
doi = {10.1007/s00259-026-08139-z},
url = {https://doi.org/10.1007/s00259-026-08139-z},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Automated joint segmentation enables quantitative molecular imaging of degeneration with [¹⁸F]FDG and [¹⁸F]NaF PET/CT
AU - William K. Lee
AU - Shiv Patil
AU - Darshil H. Patel
AU - Zhuorui Li
AU - Ondrej Fanta
AU - Om H. Gandhi
AU - Thomas J. Werner
AU - Poul F. Høilund-Carlsen
AU - Abass Alavi
PY - 2026
DA - 2026/08/21
JO - European Journal of Nuclear Medicine and Molecular Imaging
DO - 10.1007/s00259-026-08139-z
UR - https://doi.org/10.1007/s00259-026-08139-z
AB - Purpose: Manual quantification of joint metabolism and inflammation with PET is time-consuming and difficult to standardize due to ambiguous joint boundaries. We aimed to develop an automatic joint segmentation method that efficiently and reliably quantifies activity in weight-bearing joints using [18F]fluorodeoxyglucose (FDG) and sodium [18F]fluoride (NaF) PET/computed tomography (PET/CT). Methods: Segmentations were generated from CT scans using the TotalSegmentator model. From these, 30 weight-bearing joints were segmented by identifying the region between adjacent bones within a set distance threshold. Scans of healthy volunteers were analyzed (75 with FDG, 80 with NaF). Mean standardized uptake value (SUVmean) was calculated for each joint region, and Pearson correlation analysis was used to evaluate relationships with age and body mass index (BMI). Benjamini–Hochberg false discovery rate (FDR) correction was applied within each tracer–covariate family. Results: The automated approach reduced segmentation time from 30 to ~ 1.5 min per scan. Joint ROI volumes generated independently from the two PET/CT sessions of the same participant were highly concordant (median r = 0.98, median intraclass correlation coefficient 0.94). After false discovery rate correction, NaF uptake was positively correlated with BMI in the thoracic (T3-T6, T7-T8, T9-T11) and lumbosacral (L5-S1) intervertebral discs (IVDs) and in the sacroiliac (SI), hip and knee joints, and positively with age in the cervical IVDs (C1-C2, C3-C7), with negative age correlations at T4-T5, T6-T7, L1-L2 and in both SI joints (all q < 0.05). FDG uptake was positively correlated with BMI in the IVDs from C1-T6 (other than T1-T2) and T9-S1, in both SI joints, the left hip and both knees, and with age in both knees, and negatively with age at T2-T3 and in both SI joints (all q < 0.05). Associations meeting p < 0.05 but not q < 0.05 are identified in Tables 2, 3, 4 and 5. The positive cervical NaF-age associations arise in the only joints where SUVmean was appreciably correlated with ROI size and are therefore reported as provisional. Conclusion: Our segmentation approach rapidly quantifies joint metabolism and revealed significant associations between PET tracer uptake, BMI, and age. It offers a standardized framework for future PET/CT joint analysis. Clinical trial registration: Cardiovascular Molecular Calcification Assessed by 18 F-NaF PET/CT (CAMONA), ClinicalTrials.gov NCT01724749. https://clinicaltrials.gov/study/NCT01724749. Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1007/s00259-026-08139-z.
SN - 1619-7070
SN - 1619-7089
ER -