Texture analysis of ¹⁸F-FDG PET/CT for predicting the malignant nature in thymic epithelial tumors

Objectives: To investigate whether the texture analysis is useful for predicting the malignant nature in ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG)-avid thymic epithelial tumors in PET/CT.

Methods: Thirty four patients with ¹⁸F-FDG-avid thymic epithelial tumor were reviewed. The thymic epithelial tumors were classified by the WHO histological classification and Masaoka clinical staging. SUVmax and 4 textural parameters [entropy, homogeneity, intensity variability (IV) and size-zone variability (SZV)] of these tumors were obtained. Entropy reflects irregularity in the gray level co-occurrence matrix, and homogeneity measures the uniformity of the gray level co-occurrence matrix. IV measures the similarity in pixel intensities throughout the image and SZV measures the similarity in zone sizes. Heterogeneous images tend to have larger entropy, IV and SZV and lower homogeneity. These parameters were compared between the 11 low-risk (Type A, AB and B1) and 23 high-risk {Type B2, B3 and C (thymic cancer)} tumor groups and among clinical stages using the Mann-Whitney U test or Kruskal-Wallis test, and the diagnostic performance was evaluated by the area under the curve (AUC) values derived from the receiver operating characteristic analysis.

Results: The high-risk tumors showed significantly higher SUVmax (6.7 ± 2.6 vs. 5.0 ± 3.2, p=0.022), higher entropy (7.21 ± 0.32 vs. 7.04 ± 0.19, p=0.038), higher IV (42.7 ± 33.0 vs. 22.1 ± 11.7, p=0.041) and higher SZV (642.8 ± 568.8 vs. 339.6 ± 129.9, p=0.045) than low-risk tumors. There were no significant differences in SUVmax and textural parameters among clinical stages (p蠅0.05 each). The accuracy for diagnosing high-risk tumors was 67.6%, 73.5%, 67.6% and 73.5% for SUVmax, entropy, IV and SZV, respectively. No significant differences in AUC were found among these parameters (p蠅0.05 each). The accuracy of combination of either SUVmax or entropy was significantly higher than that of SUVmax alone (88.2% vs. 67.6%, p=0.045).

Conclusion: The high-risk tumors may be more heterogeneous in ¹⁸F-FDG texture features than low-risk tumors, suggesting that the texture heterogeneity parameters may be useful to predict the malignant nature of thymic epitherial tumors. Research Support: