Abstract
BACKGROUND/OBJECTIVES: Neuroendocrine tumors (NETs) of the lung account for about 30% of NETs. In localized and locally advanced forms, radical surgical resection is the standard of care. Although considered indolent tumors, they appear to be susceptible to post-surgical recurrence, with rates differing between typical and atypical carcinoid. Although still debated, several clinicopathologic factors are potentially associated with recurrence. The aim of this retrospective/prospective observational study is to evaluate the predictive role of clinicopathological factors and radiomics features in patients with NET of the lung. METHODS: From January 2021 to April 2024, 45 consecutive patients who underwent radical (R0) surgery for lung NET at the ENETS Center of Excellence of the Sant'Andrea Hospital were enrolled, all with at least 12 months of postoperative follow-up and availability of preoperative unenhanced chest CT. Clinicopathologic and radiomic factors were considered (107 radiomic features). Of the individual characteristics, the impact on recurrence was assessed by univariate logistic regression. RESULTS: Among the 45 patients included, 4 patients (8.9%) experienced disease recurrence. Among the clinicopathological features, major age at diagnosis (p = 0.020), atypical carcinoid (p = 0.010), presence of functional syndrome (p = 0.002), advanced stage at diagnosis (p = 0.013), necrosis (p = 0.017) higher Ki-67 (p = 0.001), higher mitotic count (p = 0.006), and pathologic lymph node (p = 0.006) were associated with disease recurrence. Three radiomic features were found to predict recurrence: DependenceEntropy (p = 0.049), DependenceNonUniformityNormalized (p = 0.024), and Elongation (p = 0.039). In this preliminary analysis, multivariate analysis was not performed due to the small sample size. CONCLUSIONS: This study has shown that radiomics can be a valuable tool in predicting recurrence. Currently, to our knowledge, no other studies on the possible application of radiomics as prognostic factors in patients with lung NET have been published. These encouraging findings warrant further investigations with larger, multicenter cohorts to validate these results and implement them by constructing a predictive model of recurrence.