A CT-based radiomics and deep learning signature for evaluating the somatostatin receptor 2 in non-functional pancreatic neuroendocrine tumors: A multicohort, retrospective study

#4127

Introduction: The diagnosis and treatment of Non-functional Pancreatic Neuroendocrine Tumor (NF-panNET) with unlabeled or labeled somatostatin analogues necessitate high expression of the somatostatin receptor subtype 2 (SSTR2), typically identified using PET or SPECT imaging. However, the application of PET or SPECT imaging is limited due to its low spatial resolution and unavailability in all units.

Aim(s): To develop a radiomics and deep-learning signature for predict the SSTR2 expression of NF-panNET in CT imaging.

Materials and methods: A total of 200 panNET patients from Center 1 and 50 from Center 2 underwent surgical resection, with pathology confirming SSTR2 expression. Pyradiomics extracted radiomics features, while deep-learning features extraction utilized the ResNet152 architecture. Radiomics and deep-learning signatures were developed using GBM machine-learning techniques. Diagnostic performance was evaluated using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA).

Conference:

Presenting Author:

Authors: Tang W, Wenchao G, Yinli C, Jie C,

Keywords: Non-functional Pancreatic Neuroendocrine Tumor, SSTR2, Radiomics, Deep-learning, CT,