Nuklearmedizin 2020; 59(02): 173
DOI: 10.1055/s-0040-1708372
Wissenschaftliche Poster
Radiomics
© Georg Thieme Verlag KG Stuttgart · New York

Automatic vertebra labeling using SPECT/CT images

B Barta
1   University of Szeged, Department of Computer Algorithms and Artificial Intelligence, Szeged
,
S Urbán
2   University of Szeged, Department of Nuclear Medicine, Szeged
,
Z Besenyi
2   University of Szeged, Department of Nuclear Medicine, Szeged
,
L Pávics
2   University of Szeged, Department of Nuclear Medicine, Szeged
,
N Zsótér
3   Mediso Medical Imaging Systems Ltd., Evaluation Software Department, Budapest
,
J Csirik
1   University of Szeged, Department of Computer Algorithms and Artificial Intelligence, Szeged
› Author Affiliations
Further Information

Publication History

Publication Date:
08 April 2020 (online)

Also available at
 
 

    Ziel/Aim Identification of the vertebrae is usually done manually. The automatic labeling of the vertebrae can facilitate this task by speeding up the process. However the appearance of vertebrae on low-dose CT images makes the computer-aided vertebra labeling a challenging task. On the usually used low-dose CT the separation of cervical and thoracic vertebrae can be difficult due to larger slice thickness. This work presents an automatic approach for vertebra labeling focusing on the SPECT/CT clinical protocols using low-dose CT images.

    Methodik/Methods Our method consists of two main steps. The first is the extraction of the spine from the CT volume based on automatic thresholding and Hough transformations. Then in the second step the labeling of each vertebra is done by Hessian- and rib cage analyis.

    Ergebnisse/Results The method was tested on 12 Bone SPECT/CT images (SPECT: Tc-99m-MDP, CT: 120 KeV, 50-100 mA, slice thickness: 2.5 mm) from the daily clinical routine acquired on the Mediso AnyScan SPECT/CT system. The automatic vertebra labeling succeeded in 11 cases of total 12. The results were validated by two specialists in nuclear medicine. The labeled vertebrae can be fused with the SPECT to help the localization of lesions with higher uptakes.

    Schlussfolgerungen/Conclusions The automatic vertebra labeling is feasible from low-dose CT images, which can also facilitate the localization of the lesions on SPECT images.


    #