Objective To perform external validation of a predictive model based on clinical CT imaging features for preoperative identification of high-grade patterns (HGP), such as micropapillary and solid subtypes, in early-stage lung adenocarcinoma, to guide clinical treatment decisions. Methods This study utilized a previously developed predictive model for external validation in a cohort of 650 patients with clinical stage ⅠA lung adenocarcinoma from the Fourth Hospital of Hebei Medical University. The patients in the validation cohort had an age range of 30 to 82 years, with a median age of 61 years, including 293 males (45.1%). The model incorporated factors such as tumor size, density, and lobulation features. Data analysis included the model’s discriminative ability, calibration performance, and clinical impact. Results Validation revealed that the model demonstrated good performance in discriminating HGP (area under the curve>0.7). Calibration of the original model improved its calibration performance. Decision curve analysis (DCA) indicated that the model’s predicted HGP patient population closely approximated the actual population when using a threshold probability>0.6. Conclusion This study confirms the effectiveness of a CT imaging feature-based prediction model for identifying HGP in stage ⅠA lung adenocarcinoma in a clinical setting. Successful application of this model may be significant for determining surgical strategies and improving patient prognosis. Despite certain limitations, these findings provide new directions for future research.
