Program Information
CT Texture Feature Stability Dependence On the Image Binning Approach
I Mihaylov1*, (1) Univ Miami, Miami, FL
Presentations
MO-RPM-GePD-J(B)-1 (Monday, July 31, 2017) 3:45 PM - 4:15 PM Room: Joint Imaging-Therapy ePoster Lounge - B
Purpose: To explore the effects of the binning technique on the texture features derived from gray-level co-occurrence matrices (GLCMs).
Methods: A Cathphan phantom was CT scanned seven times over a period of three weeks on Siemens Somatom 64-slice scanner. The reconstructed CT images had a resolution of 0.976x0.976x1 mm3. Four cylindrical regions of interest (ROIs) were manually outlined on the phantom for each of the seven scans. Two of those regions had fairly uniform density, while the other two had rather variable density. For each ROI four GLCMs were created – with 64 bins, with 32 bins, with fixed bin width of 4 HU, and fixed bin width of 1 HU. The GLCMs covered the entire range of HUs specific for each particular ROI. Eighteen commonly used image features were calculated form the GLCMs and the variability (defined as standard deviation divided by the average) of those features were compared between the four different GLCM representations.
Results: For the two ROIs with uniform density the variability of the fixed number of bins GLCMs was on average lower (not for all features) than the variability observed for fixed bin width GLCMs. For the other two ROIs the situation was reversed, where the fixed bin width GLCMs exhibited almost always lower variability than the fixed bin number GLCMs. The uniform density ROIs had average standard deviation of the HUs divided by the average HU values (over the seven scans) of ~1.5%, while the standard HU deviation of the more heterogeneous ROIs was in excess of 4%.
Conclusion: The uniform density ROIs present textural features which are very susceptible to machine noise. The variable density ROIs are much more realistic and therefore the use of fixed bin width GLCMs is advantageous from feature stability stand point.
Contact Email: