Predicting Preoperative GEP-NET Grade With an Automated Radiomics Model

A novel auto-segmentation model and radiomics signature may help with preoperative clinical decision making for gastroenteropancreatic neuroendocrine tumors (GEP-NETs), according to a study presented at the 2025 American Society of Clinical Oncology Gastrointestinal Cancers Symposium.

GEP-NETs have varying biologic behavior, and many prognostic factors can only be determined after pancreatectomy. Imaging analysis with radiomics may be able to show biologic data, such as tumor grade, without tissue sampling.

Dr. Pratik Chandra and colleagues from Memorial Sloan Kettering Cancer Center designed a study aimed at creating an automatic pipeline from segmentation to radiomics-signature building  to preoperatively predict tumor grade. A total of 186 patients who underwent pancreatectomy from 2003 through 2021 were included. Patients had to have adequate preoperative arterial phase computed tomography scans. They were then divided into training (n=140) and validation (n=46) cohorts.

In the training cohort, the pancreas and tumor region were manually segmented and subsequently used to train an auto-segmentation model. The validation cohort then underwent automated segmentation.

Researchers extracted 255 radiomics features from the tumor region. Correlated features were removed, and the minimum redundancy maximum relevance method was used to select the final feature set, which was used to train a support vector machine–based classifier through the training cohort to predict grade—dichotomized into grade I (n=113; 62%) versus grade II/III (n=70; 38%). Researchers then evaluated the radiomics-based prediction model in the automatically segmented validation cohort.

Results of the auto-segmentation model showed accurate segmentation of the tumor region in 72% (n=33) of patients in the validation cohort. In the training cohort, the radiomics signature produced an area under the curve (AUC) of 0.87 (0.81-0.93). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 0.94 (0.90-0.98), 0.74 (0.66-0.81), 0.72 (0.65-0.80), and 0.97 (0.94-1.00), respectively.

In the auto-segmented validation cohort (n = 33), the radiomics model produced an AUC of 0.75 (0.61-0.90); sensitivity, specificity, PPV, and NPV were 0.88 (0.77-0.99), 0.62 (0.46-0.79), 0.71 (0.56-0.87), and 0.83 (0.71-0.96), respectively.

“This automatic pipeline bypasses manual segmentation and could help incorporate a radiomics- signature into preoperative clinical decision-making for GEP-NETs,” Dr. Chandra and colleagues concluded.