Image Acquisition and Analysis
All patients underwent a 64-section contrast-enhanced CT scan (Discovery HD 750; GE Healthcare) covering the cervical region to the chest. Acquisition parameters were as follows: tube voltage, 120 kVp; section collimation, 64x1.25 mm; helical pitch, 1.375; scan time per spiral, 0.7s; image reconstruction thickness, 2.5 mm.
CT-TA was performed on the pretreatment CT examination using the commercially available TexRAD software (TexRAD Ltd). An ENT-specialized radiologist blinded to all data selecteded the CT image demonstrating the lesion largest cross-sectional area. The region of interest (ROI) was manually delineated around the tumor.
CT-TA was performed in a two-step process including image filtration and histogram quantification. Spatial scale image filtration (SSF) extracted features with different texture scales, as follows: fine (SFF=2, object radius of 2 mm), medium (SFF=3-5, 3-5 mm), coarse (SFF=6, 6mm) and without filter (SFF=0) , by using a Laplacian of Gaussian spatial band-pass filter (Figure1). Quantification of the histogram distribution within the ROI allowed the extraction of five texture parameters: mean gray-level intensity, standard deviation of the gray-level histogram distribution (SD), entropy, kurtosis, and skewness. The mean value of each texture parameters among lesions was calculated.