Purpose: Early identification of aggressive disease could improve decision support in pancreatic neuroendocrine tumor (pNET) patients prior to peptide receptor radionuclide therapy (PRRT). The prognostic value of intratumoral textural features (TF) determined by baseline somatostatin receptor (SSTR)-positron emission tomography (PET) before PRRT was analyzed. Procedures: Thirty-one patients with G1/G2 pNET were enrolled (G2, n = 23/31). Prior to PRRT with [ 177 Lu]DOTATATE (mean, 3.6 cycles), baseline SSTR-PET computed tomography was performed. By segmentation of 162 (median per patient, 5) metastases, intratumoral TF were computed. The impact of conventional PET parameters (SUV mean/max ), imaging-based TF, and clinical parameters (Ki67, CgA) for prediction of both progression-free survival (PFS) and overall survival (OS) after PRRT were evaluated. Correspondence to: Ralph Bundschuh; e-mail: ralph.bundschuh@ukbonn.deResults: Within a median follow-up of 3.7 years, tumor progression was detected in 21 patients (median, 1.5 years) and 13/31 deceased (median, 1.9 years). In ROC analysis, the TF entropy, reflecting derangement on a voxel-by-voxel level, demonstrated predictive capability for OS (cutoff = 6.7, AUC = 0.71, p = 0.02). Of note, increasing entropy could predict a longer survival (9 6.7, OS = 2.5 years, 17/31), whereas less voxel-based derangement portended inferior outcome (G 6.7, OS = 1.9 years, 14/31). These findings were supported in a G2 subanalysis (9 6.9, OS = 2.8 years, 9/23 vs. G 6.9, OS = 1.9 years, 14/23). Kaplan-Meier analysis revealed a significant distinction between high-and low-risk groups using entropy (n = 31, p G 0.05). For those patients below the ROC-derived threshold, the relative risk of death after PRRT was 2.73 (n = 31, p = 0.04). Ki67 was negatively associated with PFS (p = 0.002); however, SUV mean/max failed in prognostication (n.s.). Conclusions: In contrast to conventional PET parameters, assessment of intratumoral heterogeneity demonstrated superior prognostic performance in pNET patients undergoing PRRT. This novel PET-based strategy of outcome prediction prior to PRRT might be useful for patient risk stratification.