CT imaging-based nomogram for predicting pleural invasion in non-small cell lung cancer.

To identify risk factors for visceral pleural invasion (VPI) in non-small cell lung cancer (NSCLC) using preoperative CT characteristics and develop a nomogram for VPI prediction. We collected clinical data from 282 NSCLC patients (July 2020-June 2022) who underwent CT scans. Patients were categorized into VPI (n=70) and non-VPI (n=212) groups. Demographic, pathologic, and CT characteristics were analyzed. Logistic regression was applied to identify VPI-related determinants, and a nomogram was developed. Diagnostic performance was assessed using receiver operating characteristic (ROC) curves, decision curve analysis (DCA), and calibration curves. No significant differences were observed between the groups for gender, age, BMI, disease duration, KPS score, smoking/alcohol history, tumor location, pathologic type, or CT features (all P>0.05). The VPI group had higher proportions of moderately/poorly differentiated tumors, advanced N-stage, M1-stage, larger tumor size, pleural retraction, pleura-adherence, and irregular morphology (all P<0.05). Multivariate analysis identified significant predictors: moderate/poor differentiation (OR=2.41), N2-N3 stage (OR=3.17), M1 stage (OR=2.89), pleural retraction (OR=4.02), pleura adherence (OR=3.55), irregular morphology (OR=3.21), and larger tumor diameter (OR=1.98) (all P<0.05). The model exhibited AUCs of 0.933 (training) and 0.959 (validation). Bootstrap validation (B=1000) indicated good calibration. The VPI group had a lower 3-year survival rate (69.57%) compared to the non-VPI group (80.66%) (P<0.05). VPI in NSCLC is associated with worse prognosis. The CT features, including pleural indentation and irregular morphology, may be used to predict VPI. The nomogram developed will be a valuable tool for clinical decision-making and improving patient outcomes.