Radiomics in advanced gastroenteropancreatic neuroendocrine neoplasms: Identifying responders to somatostatin analogs

• Published in: Journal of neuroendocrinology p. e13472
• Main Authors: Polici, Michela, Caruso, Damiano, Masci, Benedetta, Marasco, Matteo, Valanzuolo, Daniela, Dell'Unto, Elisabetta, Zerunian, Marta, Campana, Davide, De Santis, Domenico, Lamberti, Giuseppe, Iannicelli, Elsa, Prosperi, Daniela, Annibale, Bruno, Laghi, Andrea, Panzuto, Francesco, Rinzivillo, Maria
• Format: Journal Article
• Language: English
• Published: 20-11-2024
• ISSN: 0953-8194; 1365-2826; 1365-2826
• DOI: 10.1111/jne.13472

To evaluate a radiomic strategy for predicting progression in advanced gastroenteropancreatic neuroendocrine tumor (GEP‐NET) patients treated with somatostatin analogs (SSAs). Fifty‐eight patients with GEP‐NETs and liver metastases, with baseline computerized tomography (CT) scans from June 2013 to November 2020, were studied retrospectively. Data collected included progression‐free survival (PFS), overall survival (OS), tumor grading, death, and Ki67 index. Patients were categorized into progressive and non‐progressive groups. Two radiologists performed 3D liver segmentation on baseline CT scans using 3DSlicer v4.10.2. One hundred six radiomic features were extracted and analyzed ( T ‐test or Mann–Whitney). Radiomic feature efficacy was evaluated via receiver operating characteristic curves, and both univariate and multivariate logistic regression were used to develop predictive models. A significance level of p < .05 was maintained. Of 55 patients, 38 were progressive (median PFS and OS: 14 and 34 months, respectively), and 17 were non‐progressive (median PFS and OS: 58 months each). Six radiomic features significantly differed between groups ( p < .05), with an area under the curve (AUC) range of 0.64–0.74. Ki67 was the only clinical parameter significantly associated with progression risk (odds ratio (OR) = 1.14, p < .05). The combined radiomic features and Ki67 model proved most effective, showing an AUC of 0.814 ( p = .008). The radiomic model alone did not reach statistical significance ( p = .07). A combined model incorporating radiomic features and the Ki67 index effectively predicts disease progression in GEP‐NET patients eligible for SSA treatment.