• GEPNETs
• PET/TC
• radiomic
• texture features

GEPNETs represent the most common NET subtype, comprising 55–70% of all NETs, they include carcinoid tumors of the gastrointestinal tract and pancreatic NETs, overall incidence of NET is less than six new cases per year per 100,000 individuals therefore they are considered rare neoplasms, but their incidence and prevalence continue to rise globally.
In this study we have selected only patients affected by GEPNET G2 (Ki 67 3/20%) recruited during the interval between 19/04/2019 and 9/06/2022. The primary aim is to evaluate the disease’s heterogeneity using [68 Ga] DOTA-TOC PET/CT multiparametric imaging both through a traditional approach (qualitative, semi-quantitative) and through a quantitative approach, using texture features analysis.
The study cohort included 49 patients, 24 women and 25 men, mean age 60,92 years, median 64,95, Std. dev. 15,75 (range 20,94-85,73). A total of 75 [68 Ga] DOTA TOC PET/CT examinations were collected; 50 were baseline scans while 25 scans were follow-up [68 Ga] DOTA-TOC PET/CT examinations.
All [68 Ga] DOTA-TOC PET/CT scans were analyzed using the EANM guidelines, the acquisition comprises a whole-body scan (from the head to middle of the upper leg), images were acquired about 30/40 minutes after [68 Ga] DOTATOC administration (2mBq per kg).
[68 Ga] DOTATOC PET/CT was positive in 60 cases, and it was negative in 15 cases. 158 lesions were identified divided into 136 metastases (79 liver, 43 lymph nodes, 6 bones, 4 peritoneal metastases, 1 lung and 1 heart) and 22 primary tumors (16 pancreatic lesions 4 ileal lesions, 1 rectum, 1 stomach). Using LifeX software (http://www.lifexsoft.org) all 158 lesions were segmented and from all VOI generated texture features were extracted.

At univariate analysis we evaluated the correlation between Ki-67 value, Chromogranin value and clinical stage, primary tumor and with texture features, showing Chromogranin as a good biomarker for disease clinical stage.
Starting from the texture features, we also implemented a multivariate analysis (LDA linear discriminant analysis) to create models useful for predicting the stage of the disease, the site of the primary tumor and to characterize the lesions based on their nature (T, N, M), allowing us to evaluate the heterogeneity of GEPNETs G2 with a quantitative approach. We obtained, in discriminating the nature of the lesions an AUC respectively of 0.78 for T, 0.87 for M and 0.89 for N.Conclusions The radiomic features extracted from [68 Ga] DOTATOC PET / CT are able to discriminate negative patients and the nature of the lesions. The heterogeneity study shows that there are two disease phenotypes based on significant texture features, suggesting, in agreement with the literature, a different response to PRRT.