Radiomic analysis performed on 68Ga-DOTANOC PET/CT allows discriminating primary pancreatic neuroendocrine tumour grading

Purpose or Learning Objective To evaluate the effectiveness of radiomic features (RFs) computed on 68Ga-DOTANOC PET/CT to discriminate grade 1 (G1) and grade 2 (G2) pancreatic neuroendocrine tumours (panNET). Methods or Background Twenty-eight patients with primary panNET (M:F=13:15; mean age: 56 years old [17-78]; G1=13, G2=15) whose grading was assessed after surgical excision were enrolled in this study. First and second-order RFs were computed on the standardized uptake value (SUV) maps pertaining to the whole lesion. 1770 coupled texture RFs initially were considered, but those with a high correlation were subsequently discarded. The linear discriminant analysis was used on the remaining couples to build as many discriminative radiomic models (RMs), and their discrimination capability was assessed using two-tail Wilcoxon rank-sum test on the RF space and the Bonferroni correction (p-value=0.001). The Area under the curve (AUC) of the receiver operating characteristic (ROC) was used to assess RM performances, together with sensitivity, specificity and accuracy. The best RM was selected as that showing the lowest p-value and the highest AUC. Results or Findings The RF couple made of first-order skewness and second-order sum entropy shows the highest significant performance (p-value=0.0003) with sensitivity=93%, specificity=85%, accuracy=89%, achieved with 2 false positive and 1 false negative. Conclusion The radiomic model identified shows a high statistical significance in discriminating G2 from G1 primary panNET and it is promising for the validation step. This preliminary study suggests that the texture-based RFs computed on 68Ga-DOTANOC PET/CT could represent a valid tool for tumour grade assessment, potentially useful in panNET not amenable to biopsy before surgery. Limitations The limited number of sample size.