165P {\ast} {PREOPERATIVE} {POSITRON} {EMISSION} {TOMOGRAPHY} {FRACTAL} {BIOPSY} {OF} {THYMIC} {EPITHELIAL} {NEOPLASM}

Aim: To determine whether fractal analysis of FDG PET uptake, that quantifies tumor complexity and heterogeneity, helps or not preoperative identification of Low Risk (LR) Thymic Epithelial Neoplasm (TEN) & High Risk (HR) TEN. Methods: Retrospective cohort of 26 consecutive patients with pathologically proven TEN, categorized according to WHO classification. PET Standardized Uptake Value Tumor/Mediastinum (SUV T/M) ratio was calculated. Fractal geometry, characterized by relationship between measure (M), scale (ε), scaling constant (k), self-affinity dimension (D), was used: M(ε) = kε-D. Cutoffs were ruler ε, voxels (radioactivity > cutoff) were M(ε), pixels were D. Morphological Fractal Dimension (m-FD), quantitative index of morphological complexity, and density Fractal Dimension (d-FD), expression of FDG distribution heterogeneity, were calculated. Comparison of SUV T/M, m-FD, d-FD between LR TEN & HR TEN was performed with unpaired t – test. Pearson's correlation coefficients evaluated relationships. Values plotted nearest upper left corner from Receiver Operating Characteristics (ROC) analyses indicated best diagnostic accuracy. Sensitivity, specificity, accuracy were calculated, using appropriate cutoffs. Diagnostic accuracy was compared using Areas under ROC Curves (AUC); ROC curves were compared using critical z test. Significance level was set to 0.05. Results: SUV T/M ratio and d-FD highly correlate with stage (SUV T/M: r = 0.71, p = 0.0001; d-FD: r = 0.70, p = 0.0001) and classes of risk (SUV T/M: r = 0.83, p < 0.0001; d-FD: r = 0.76, p < 0.0001). Cutoff values were: SUV T/M = 2.63, m-FD = 0.031, d-FD = 0.584. Combined use of PET parameters for differential diagnosis was examined: pooled d-FD with SUV T/M improved accuracy to 89%. Distribution of PET parameters WHO Classification SUV T/M m-DF d-DF - Type a (#3, 11.5%) 1.82 ± 0.63 0.024 ± 0.004 0.099 ± 0.091 - Type ab (#11, 42.3%) 1.86 ± 0.43 0.029 ± 0.007 0.256 ± 0.247 - Type b1 (#5, 19.3%) 1.99 ± 0.49 0.030 ± 0.011 0.246 ± 0.399 - Type b2 (#3, 11.5%) 4.33 ± 0.91 0.036 ± 0.001 0.979 ± 0.176 - Type b3 (#4, 15.4%) 3.59 ± 1.05 0.037 ± 0.014 0.793 ± 0.163 Classes of Risk * SUV T/M m-DF d-DF - Low Risk (#19, 73.1%) 1.89 ± 0.45 0.028 ± 0.008 0.228 ± 0.251 - High Risk (#7, 26.9%) 3.91 ± 0.99 0.037 ± 0.012 0.872 ± 0.183 Comparison of Diagnostic Ability SUV T/M m-DF d-DF - AUC 0.974 0.737 0.947 - Cutoff 2.63 0.031 0.584 - Sensitivity (%) 92.11 42.11 89.47 - Specificity (%) 85.71 57.14 78.57 - Accuracy (%) 84.62 53.85 80.77 * Low Risk included thymoma types a, ab and b1; High Risk included thymoma types b2 and b3 Conclusions: d-FD and SUV T/M provided information useful to predict histology and classes of risk. Heterogeneity of PET FDG uptake evaluated by fractal analysis was useful to discriminate preoperatively LR TEN from HR TEN. Disclosure: All authors have declared no conflicts of interest.