The value of a nomogram model based on CT imaging features in differentiating duodenal gastrointestinal stromal tumors from pancreatic head neuroendocrine tumors.

Objective: To construct a nomogram model based on multi-slice spiral CT imaging features to predict and differentiate between duodenal gastrointestinal stromal tumors (GISTs) and pancreatic head neuroendocrine tumors (NENs), providing imaging evidence for clinical treatment decisions.

Methods: A retrospective collection of clinical information, pathological results, and imaging data was conducted on 115 cases of duodenal GISTs and 76 cases of pancreatic head NENs confirmed by surgical pathology at Zhongshan Hospital Fudan University from November 2013 to November 2022. Comparative analysis was performed on the tumor's maximum diameter, shortest diameter, long diameter/short diameter ratio, tumor morphology, tumor border, central position of the lesion, lesion long-axis direction, the relationship between tumor and common bile duct (CBD), duodenal side ulceration of the lesion, calcification, cystic and solid proportion within the tumor, thickened feeding arteries, tumor neovascularization, distant metastasis, and CT values during plain and enhanced scans in arterial and venous phases. Statistical analysis was conducted using t-tests, Mann-Whitney U tests, and χ² tests. Univariate and multivariate logistic regression analyses were used to identify independent predictors for differentiating duodenal GISTs from pancreatic head NENs. Based on these independent predictors, a nomogram model was constructed, and the receiver operating characteristic (ROC) curve was used to evaluate the diagnostic performance of the model. The nomogram was validated using a calibration curve, and decision curve analysis was applied to assess the clinical application value of the nomogram.

Results: There were significant differences in the duodenal GISTs group and the pancreatic head NENs group in terms of longest diameter (P < 0.001), shortest diameter (P < 0.001), plain CT value (P < 0.001), arterial phase CT value (P < 0.001), venous phase CT value (P = 0.002), lesion long-axis direction (P < 0.001), central position of the lesion (P < 0.001), the relationship between tumor and CBD(< 0.001), border (P = 0.004), calcification (P = 0.017), and distant metastasis (P = 0.018). Multivariate logistic regression analysis identified uncertain location (OR 0.040, 95% CI 0.003-0.549), near the duodenum (OR 0, 95% CI 0-0.009), with the lesion long-axis direction along the pancreas as a reference, along the duodenum (OR 0.106, 95% CI 0.010-1.156) or no significant difference (OR 4.946, 95% CI 0.453-54.017), and the relationship between tumor and CBD (OR 0.013, 95% CI 0.001-0.180), shortest diameter (OR 0.705, 95% CI 0.546-0.909), and calcification (OR 18.638, 95% CI 1.316-263.878) as independent risk factors for differentiating between duodenal GISTs and pancreatic head NENs (all P values < 0.05). The combined diagnostic model's AUC values based on central position of the lesion, calcification, lesion long axis orientation, the relationship between tumor and CBD, shortest diameter, and the joint diagnostic model were 0.937 (0.902-0.972), 0.700(0.624-0.776), 0.717(0.631-0.802), 0.559 (0.473-0.644), 0.680 (0.603-0.758), and 0.991(0.982-0.999), respectively, with a sensitivity of 97.3% and a specificity of 93.0% for the joint diagnostic model. The nomogram model's AUC value was 0.985(0.973-0.996), with a sensitivity and specificity of 94.7% and 93.9%, respectively. The calibration curve indicated good agreement between predicted and actual risks. Decision curve analysis verified the clinical application value of the nomogram.

Conclusion: The nomogram model based on CT imaging features effectively differentiates between duodenal GISTs and pancreatic head NENs, aiding in more precise clinical treatment decisions.