Non-linear image blending improves visualization of head and neck primary squamous cell carcinoma compared to linear blending in dual-energy CT

Abstract

To compare non-linear and linear image-blending post-processing techniques in dual-energy CT (DECT) of primary head and neck squamous cell carcinoma (SCC) regarding subjective and objective image quality. Head and neck DECT studies from 69 patients (48 male, 21 female; mean age 62.3 years) were retrospectively evaluated. All tumour lesions were histologically confirmed SCC. Linearly blended 80/140kVp images series with varying weighting factors of 0.3 (M_0.3), 0.6 and 0.8 were compared with non-linearly blended images. Attenuation of tumour lesion, various soft-tissue structures, the internal jugular vein, and image noise were measured, tumour signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. Overall image quality, delineation of tumour lesion, image sharpness, and noise level were rated individually by three radiologists using five-point Likert scales. Interobserver agreement was calculated using intraclass correlation coefficient (ICC). Enhancement of tumour lesions (non-linear, 137.5±20.1HU; M_0.3, 92.7±14.4HU; M_0.6, 110±15.4HU; M_0.8, 123±18.2HU), CNR (non-linear, 12±8; M_0.3, 4±4.7; M_0.6, 7.5±5.5; M_0.8, 8±5.5), subjective overall image quality and tumour delineation were significantly increased (all p<0.001) with the non-linear blending technique compared to all investigated linear blending weighting factors. Overall interobserver agreement was substantial (ICC 0.70; 95% CI: 0.66-0.73). Post-processing of DECT using a non-linear blending technique provides improved objective and subjective image quality of head and neck SCC compared to linearly blended images series.