MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Impact of X-Ray Beam Quality On CT Radiomic Features Across Multiple Scanners

M Shafiq ul Hassan1,2*, G Zhang1,2 , K Latifi1 , G Ullah2 , R Gillies1 , E Moros1,2 , (1) H Lee Moffitt Cancer Center and Research Institute, Tampa, FL, (2) University of South Florida, Tampa, FL,

Presentations

(Sunday, 7/29/2018) 4:00 PM - 4:55 PM

Room: Room 205

Purpose: Recent studies suggested that radiomic features are significantly affected by CT imaging parameters. X-ray tube voltage (kVp, i.e., quality or energy) is one of the CT acquisition parameters commonly varied in CT imaging. We investigated the impact of kVp on CT radiomic features using cylindrical texture inserts.

Methods: Three texture inserts made of shredded rubber, sycamore wood and a composite of rubber, wood and airsoft balls (i.e., heterogeneous) were used in conjunction with a Gammex 467 phantom (G467). Scans of the inserts within the G467 phantom were acquired on 4 CT scanners. Scans with kVp settings of 80, 100, 120 & 140 were acquired while keeping all other parameters constant. A spherical ROI was contoured on the images of the inserts and 80 radiomic features were extracted. The absolute value of the coefficient of variation (COV) was calculated for all features. The variability (%COV) results were classified in three groups: very small (%COV < 10%), intermediate (10% ≤ %COV ≤ 20%) and large variability (%COV>20%).

Results: More than 70% of the features had %COV < 20% for rubber and heterogeneous inserts across all scanners. Most of the texture features and 40% of all features had large variability (%COV>20%) for wood insert across all scanners. Shape features were the most robust as expected. Intensity and GLCM feature groups were second and third in robustness, with 80% and 70%, respectively, of all features with %COV ≤ 20%. Intensity-skewness and Intensity-kurtosis were sensitive to kVp across all scanners. The variability in GLRLM features was scanner dependent. Most GLSZM features and NGTDM-busyness were sensitive to kVp across all scanners.

Conclusion: Overall, texture features were sensitive to kVp for the homogenous wood insert across all scanners. However, texture features, except GLSZM, were robust to kVp for the composite and rubber inserts across all scanners.

Funding Support, Disclosures, and Conflict of Interest: Research project is partly supported by NIH/NCI grant RO1-CA190105-01

Keywords

Phantoms, Quantitative Imaging, CT

Taxonomy

IM- CT: Phantoms - physical

Contact Email