Click here to


Are you sure ?

Yes, do it No, cancel

Multi-Energy Study of Impact of CT Hounsfield Unit Range in Gray Level Discretization On Radiomic Feature Stability

A Chatterjee1*, M Vallieres1 , G Romero-Sanchez2 , A Perez-Lara2 , R Forghani2 , J Seuntjens1 , (1) McGill University, Medical Physics Unit, Montreal, (2) McGill University, Department of Radiology, Montreal


(Monday, 7/15/2019) 9:30 AM - 10:00 AM

Room: Exhibit Hall | Forum 5

Purpose: Texture feature calculation requires discretization of image intensities within the region-of-interest. Two discretization approaches are: a fixed number of bins (FBN), or bins of a fixed bin size (FBS). A crucial choice is the voxel intensity range used for binning. We aimed to assess the effect of this choice on radiomic features.

Methods: The dataset comprised 95 patients with head-and-neck squamous-cell-carcinoma. The dual energy CT data was reconstructed at 21 electron energies (40, 45… 140 keV). Each texture feature was calculated 64 ways (4 voxel sizes, 4 binning algorithms, 4 gray level discretizations). All features were calculated five times: original choice, left shift (-10/-20 HU), right shift (+10/+20 HU). For each feature, Spearman correlation between nominal and four variant HU ranges (±10 HU, ± 20 HU) were calculated to determine feature stability. This was done for six texture feature types (GLCM, GLRLM, GLSZM, GLDZM, NGTDM, and NGLDM) separately. This analysis was repeated for the four binning algorithms (FBN, FBNequal, FBS, FBSequal).

Results: The chosen window was -50 to 350 HU. FBN and FBNequal algorithms showed good stability (correlation values consistently above 0.9). For FBS and FBSequal algorithms, while median values exceeded 0.9, the 95% lower bound decreased as a function of energy, with poor performance over the entire spectrum. FBNequal was the most stable algorithm, and FBS the least. For FBS, no texture type has good stability; GLRLM performed best, and GLCM the worst (reaching negative correlation values). For FBSequal, no texture type has good stability; GLRLM was the winner (good stability for ±10 HU but not ±20 HU); least stable were GLDZM and GLSZM.

Conclusion: We scrutinized how the choice of HU window used to discretize CT data affects radiomic features. Future analyses should account for this source of uncertainty when evaluating the robustness of their radiomic signature.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by CREATE Medical Physics Research Training Network grant of the Natural Sciences and Engineering Research Council (Grant number: 432290), and the Canadian Institutes of Health Research Foundation Grant FDN-143257.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email