Texture analysis (TA) applied to CT imaging is an intensely studied topic and many studies suggested TA potential value in imaging characterisation for diagnostic purposes in different fields. However, often authors do not consider the reproducibility and the robustness versus variations in acquisition parameters; in this work, we wanted to explore the robustness of the TA features extracted from CT images. We scanned a commercial phantom (CIRS model 062M) containing plugs with nine different tissue equivalent electron densities using two different CT scanners of the same vendor and changing tube current (100 and 200 mA without modulation) and peak voltage (80 and 140 kVp). After the segmentation, we extracted TA features with LifeX and data were then statistically analysed using the generic estimate equations (GEE) method. Our results suggest that only seven out of 37 TA features extracted are not affected by variation in acquisition parameters considered in this study: GLRLM lgre, GLRLM srlge, GLRLM lrgle, GLZLM lze, GLZLM lgze, GLZLM szlge, GLZLM lzlge. Definitively, we highlighted the importance of a careful study of the dependence of TA parameters on acquisition modalities and analysis before their application in clinical studies.

Effects of CT FOV displacement and acquisition parameters variation on texture analysis features

Nardone V;
2018

Abstract

Texture analysis (TA) applied to CT imaging is an intensely studied topic and many studies suggested TA potential value in imaging characterisation for diagnostic purposes in different fields. However, often authors do not consider the reproducibility and the robustness versus variations in acquisition parameters; in this work, we wanted to explore the robustness of the TA features extracted from CT images. We scanned a commercial phantom (CIRS model 062M) containing plugs with nine different tissue equivalent electron densities using two different CT scanners of the same vendor and changing tube current (100 and 200 mA without modulation) and peak voltage (80 and 140 kVp). After the segmentation, we extracted TA features with LifeX and data were then statistically analysed using the generic estimate equations (GEE) method. Our results suggest that only seven out of 37 TA features extracted are not affected by variation in acquisition parameters considered in this study: GLRLM lgre, GLRLM srlge, GLRLM lrgle, GLZLM lze, GLZLM lgze, GLZLM szlge, GLZLM lzlge. Definitively, we highlighted the importance of a careful study of the dependence of TA parameters on acquisition modalities and analysis before their application in clinical studies.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11591/462867
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 5
  • ???jsp.display-item.citation.isi??? 4
social impact