Abstract
Accurate detection of the histological grade of pancreatic neuroendocrine tumors (PanNETs) is important for patients' prognoses and treatment. Traditional grading methods based on mitotic count and Ki-67 index are subjective and time-consuming. In this work, we developed and validated a pathomics model for accurate grading of PanNETs. Pathomics features were extracted from H&E whole slide images (WSIs) using Lasso regression to create a pathomics score. Its performance was evaluated in three cohorts involving 2 centers and 272 patients. This score was significantly associated with PanNET grade and could differentiate between high- and low-risk groups. In the validation and test cohorts, the pathomics model, which combined the pathomics score with clinical features, achieved AUCs of 0.85 and 0.93, respectively. In conclusion, our model enhances data processing efficiency and provides a quantitative assessment of histological features, holding promise for guiding individualized treatment and risk stratification in PanNET management.This study was registered at ClinicalTrials. Trial number: gov ChiCTR2400090898.