CT spectral curve in differentiating spinal tumor metastasis and infections

Abstract

OBJECTIVE To evaluate the value of CT spectral curve in differentiating spinal tumor metastasis (STM) from spinal infections (SI). METHODS In the study, 29 STM and 18 SI patients proved pathologically and clinically were examined by dual energy spectral CT (DESCT). The monochromatic images and CT spectral curves were generated automatically by GSI Viewer software. The attenuation values at different energy levels (40-140 keV, every 10 keV), the attenuation values of the lesions on the conventional polychromatic CT images and the gradients of the curve were calculated and compared between STM and SI. RESULTS The median age of STM and SI (58 years vs. 64 years) were not significantly different (U=171, P=0.4). The attenuation values of STM at 40-100 keV were 281.79 (143.67, 446.19) HU, 199.68 (100.04, 321.49) HU, 151.54 (81.47, 243.49) HU, (122.64±27.72) HU, (99.90±23.88) HU, (85.82±21.61) HU, and (75.94±20.27) HU, respectively, which were significantly higher than SI: 185.29 (164.19, 277.03) HU, 138.44 (124.98, 238.56) HU, 105.46 (92.94, 169.53) HU, (93.77±15.55) HU, (79.15±12.84) HU, (68.99±11.75) HU, and (62.22±11.71) HU (all P < 0.05). The attenuation values at 110-140 keV and the attenuation value on the conventional CT images were not significantly different between STM and SI. The gradient of CT spectral curve of STM was 2.43±0.58, which was higher than the value of 1.50±0.40 for SI (P < 0.001). Using 1.72 and 248.80 HU as the threshold value for CT spectral curve slope and the attenuation value at 40 keV, could obtain the area under receiver operating characteristic (ROC) curve of 0.905 and 0.892, sensitivity of 88.0% and 80.0%, and specificity of 76.9% and 92.3%. CONCLUSION CT spectral curve provides valuable semi-quantitative information for the differential diagnosis of STM and SI, which can be used as a supplement to traditional CT imaging.