Prediction of non-functioning pancreatic neuroendocrine tumor grades with fractal analysis of preoperative contrast-enhanced computed tomography images

Abstract

PurposeTo assess intra-tumoral heterogeneity (ITH) via fractal analysis of preoperative contrast-enhanced computed tomography (CT) images to predict pathological grades in non-functioning pancreatic neuroendocrine tumors (NF-PNETs) and verify its impact on patient survival. MethodsThis retrospective study enrolled 40 patients with NF-PNET resected in our institution during a period from July 2005 to December 2018, except functioning tumors, unidentified tumors in CT, patients without preoperative contrast-enhanced CT. CT images were analyzed using plugin software for calculating fractal dimension (FD), and the maximum value was denoted as “FDmax,” and compared with pathological grades and patient survival between G1 and G2/3 group separating according to two different Ki-67 index thresholds (3% and 5%). All CT images were acquired in three-phases and arterial phase images were examined. ResultsKi-67 index and FDmax showed a direct correlation with significance (p < 0.01). The mean FDmax of the G2/3 tumor group was significantly higher than that of the G1 tumor group (p < 0.01 in both 3% and 5% thresholds). In the ROC analysis, FDmax showed 0.773 of AUC, and cut-off value of 1.036 reported 62.5 % sensitivity, 90.0 % specificity, 86.2 % PPV, and 70.6 % NPV to distinguish G2/3 patients. The high-FD (≥1.036) group showed a significantly shorter disease-free survival (DFS) than the low-FD group (p = 0.0128). In multivariate analysis of prognostic factors, high FD was the only significant factor for DFS (HR, 5.793; 95 % CI: 1.213−27.664; p = 0.028). ConclusionsThe tumor’s FDmax using CE-CT analysis might be a potential biomarker for preoperative prediction of G2/3 tumors, and predicting recurrence.