Publikation

First performance evaluation of software for automatic segmentation, labeling and reformation of anatomical aligned axial images of the thoracolumbar spine at CT

Wissenschaftlicher Artikel/Review - 13.12.2014

Bereiche
PubMed
DOI

Zitation
Scholtz J, Wichmann J, Kaup M, Fischer S, Kerl J, Lehnert T, Vogl T, Bauer R. First performance evaluation of software for automatic segmentation, labeling and reformation of anatomical aligned axial images of the thoracolumbar spine at CT. Eur J Radiol 2014; 84:437-42.
Art
Wissenschaftlicher Artikel/Review (Englisch)
Zeitschrift
Eur J Radiol 2014; 84
Veröffentlichungsdatum
13.12.2014
eISSN (Online)
1872-7727
Seiten
437-42
Kurzbeschreibung/Zielsetzung

OBJECTIVES
To evaluate software for automatic segmentation, labeling and reformation of anatomical aligned axial images of the thoracolumbar spine on CT in terms of accuracy, potential for time savings and workflow improvement.

MATERIAL AND METHODS
77 patients (28 women, 49 men, mean age 65.3±14.4 years) with known or suspected spinal disorders (degenerative spine disease n=32; disc herniation n=36; traumatic vertebral fractures n=9) underwent 64-slice MDCT with thin-slab reconstruction. Time for automatic labeling of the thoracolumbar spine and reconstruction of double-angulated axial images of the pathological vertebrae was compared with manually performed reconstruction of anatomical aligned axial images. Reformatted images of both reconstruction methods were assessed by two observers regarding accuracy of symmetric depiction of anatomical structures.

RESULTS
In 33 cases double-angulated axial images were created in 1 vertebra, in 28 cases in 2 vertebrae and in 16 cases in 3 vertebrae. Correct automatic labeling was achieved in 72 of 77 patients (93.5%). Errors could be manually corrected in 4 cases. Automatic labeling required 1min in average. In cases where anatomical aligned axial images of 1 vertebra were created, reconstructions made by hand were significantly faster (p<0.05). Automatic reconstruction was time-saving in cases of 2 and more vertebrae (p<0.05). Both reconstruction methods revealed good image quality with excellent inter-observer agreement.

CONCLUSION
The evaluated software for automatic labeling and anatomically aligned, double-angulated axial image reconstruction of the thoracolumbar spine on CT is time-saving when reconstructions of 2 and more vertebrae are performed. Checking results of automatic labeling is necessary to prevent errors in labeling.