Noninvasive evaluation of hypoxia in rabbit VX2 lung transplant tumors using spectral CT parameters and texture analysis.

AIM

Noninvasive evaluation of hypoxia in rabbit VX2 lung transplant tumors using spectral CT parameters and texture analysis.

MATERIALS AND METHODS

Twenty-five VX2 lung transplant tumors of twenty-two rabbits were included in the study. Contrast-enhanced spectral CT scanning in the arterial phase (AP) and venous phase (VP) was performed. Tumors were divided into strong and weak hypoxic groups by hypoxic probe staining results. Spectral CT image-related parameters [70 keV CT value, normalized iodine concentration (NIC), slope of spectral HU curve (λ)] were measured and the texture analysis on the monochromatic images was performed. Imaging parameters and texture features between tumors with different hypoxic degrees were compared and their diagnostic efficacies for predicting hypoxia in lung cancers were analyzed using receiver operating characteristic (ROC) curve.

RESULTS

NIC in VP and λ in VP of the strong hypoxic group were significantly higher than those in the weak hypoxic group (p < 0.05). For the texture features, entropy in VP and kurtosis in AP were significantly different between the two hypoxic groups. According to ROC analysis, λ in VP had a better diagnostic ability for predicting hypoxia in tumors [Area Under Curve (AUC): 0.883, sensitivity: 85.7%, specificity: 100%]. The combination of four features improved AUC to 0.955.

CONCLUSION

NIC in VP, λ in VP, entropy in VP and kurtosis in AP have certain values in predicting tumor hypoxia and a combination of image parameters and texture features improves diagnostic efficiency.