Automated evaluation of normal uptake in different skeletal parts using 18F-sodium fluoride (NaF) PET/CT using a new convolutional neural network method
Journal article, 2017
Material and Methods: Patients with biopsy-verified high-risk prostate cancer and a negative or inconclusive bone scintigraphy and no metastatic lesions on 18F-NaF PET/CT (performed March 2008 - June 2010) were retrospectively included (n=48). Whole-body PET scans were acquired 1-1.5 h after i.v. injection of 4 MBq/kg 18F-NaF (max 400 MBq). CT scans were obtained immediately after the PET scan. Thoracic and lumbar vertebrae, sacrum,
pelvis, ribs, scapulae, clavicles and sternum were automatically segmented in the CT images, using a method based on a convolutional neural network, to obtain the volume of each skeletal region. The network was trained using a separate group of CT scans with manual segmentations. Mean and maximum SUV (SUVmean and SUVmax) were subsequently measured for each skeletal part in the PET scans.
Results: Average (SD) SUVmean for the skeletal regions were: Thoracic vertebrae 0.98 (0.20), lumbar vertebrae 0.96 (0.19), sacrum 0.75 (0.15), pelvis 0.73 (0.16), ribs 0.41 (0.11), scapulae 0.46 (0.11), clavicles 0.50 (0.16) and sternum 0.61 (0.13). Average (SD) SUVmax for the skeletal regions were: Thoracic vertebrae 1.95 (0.66), lumbar vertebrae 2.10 (0.78), sacrum 2.22 (0.77), pelvis 1.99 (0.82), ribs 1.19 (0.35), scapulae 1.94 (0.98), clavicles 2.00 (1.03) and sternum 1.68 (0.44).
Conclusion: We present a new method to segment and quantify uptake in skeletal regions in 18F-NaF PET/CT. Various parts of the bone have different SUVs in patients with regional prostate cancer. Vertebrae and pelvis have higher SUVs than ribs. The highest SUVmax were found in the thoracic and lumbar vertebrae. The findings are of importance for interpretation of 18F-NaF PET/CT.
Author
May Sadik
Sahlgrenska University Hospital
Sarah Lindgren Belal
Lund University
Reza Kaboteh
Sahlgrenska University Hospital
Olof Enqvist
Chalmers, Signals and Systems, Signal Processing and Biomedical Engineering
Johannes Ulén
Eigenvision AB
Henrik Kjölhede
Lund University
Ola Bratt
Lund University
Lars Edenbrandt
Sahlgrenska University Hospital
Elin Trägårdh
Lund University
European Journal of Nuclear Medicine and Molecular Imaging
1619-7070 (ISSN) 1619-7089 (eISSN)
Vol. 44 Supplement 2Subject Categories
Radiology, Nuclear Medicine and Medical Imaging
Medical Image Processing