What question did this study set out to answer?

To evaluate the predictive value of CT features for visceral pleural invasion in early-stage lung adenocarcinoma with pleural contact.

April 3, 2026Cancer Imaging0 citationsOpen Access

CT predictors of visceral pleural invasion in subsolid nodular pulmonary adenocarcinoma: differences between direct and indirect tumor-pleura contact

YWY Wang DLDeng Lyu XDXiaoli Deng

Key Points

To evaluate the predictive value of CT features for visceral pleural invasion in early-stage lung adenocarcinoma with pleural contact.
Retrospective analysis of 464 patients across three hospitals
Divided dataset into training (n=244) and validation (n=119) cohorts
Performed multivariate logistic regression to identify optimal predictor variables
Developed a predictive model and nomogram, validated on an external cohort (n=101)
Classified tumor-pleura contact into direct and indirect types and analyzed CT features accordingly.
Identified independent risk factors for VPI, including solid component diameter and tumor-pleura relationship types
AUC values for the predictive model were 0.833, 0.781, and 0.796 across the cohorts
Solid component diameter and specific tumor-pleura relationship types showed significant differences between direct and indirect contact types.
Developed model demonstrated acceptable diagnostic efficacy for predicting VPI.

Abstract

Preoperative accurate prediction of visceral pleural invasion (VPI) in subsolid nodular pulmonary adenocarcinoma (PA) can provide guidance for the surgical method of early lung cancer. The aim of this study is to explore the value of CT features in predicting VPI in clinical stage IA subsolid nodular peripheral PA with contact with the pleura. A total of 464 patients across three hospitals were retrospectively analyzed. The internal dataset (n = 363) was divided into training cohort (n = 244) and internal validation cohort (n = 119) based on examination time, using a 7:3 ratio. Patients were categorized into VPI-positive and VPI-negative groups according to the pathological diagnosis. The CT features were analyzed, including pleural end characteristics, tumor signs, and tumor-pleural features. Multivariate logistic regression analyses were performed to screen the optimal combination of variables for predicting VPI. Based on these variables, a predictive model and corresponding nomogram were developed, and their predictive performance was evaluated using external validation cohort (n = 101). According to the type of tumor-pleural contact, it was divided into indirect and direct contact types. Subgroup analyses were subsequently performed for each category to identify CT features indicative of VPI. Multivariable logistic regression analysis identified the optimal combination of predictor variables, including solid component diameter, tumor-pleura relationship, vacuole sign, vascular convergence sign, and soft tissue density shadow of pleural end. Among them, solid component diameter (OR = 1.17), the Type III classification of the tumor-pleura relationship (OR = 7.43), Type IV (OR = 37.22), and Type V (OR = 13.62) were independent risk factors. The AUC values of the model for predicting VPI in three cohorts were 0.833, 0.781, and 0.796, respectively. Subgroup analysis further revealed that solid component diameter (OR = 1.12), Type IV tumor-pleura relationship (OR = 27.63), vascular convergence sign (OR = 10.68) and soft tissue density shadow of pleural end (OR = 10.48) were independent risk factors for indirect contact type, whereas solid component diameter (OR = 1.15), Type V tumor-pleura relationship (OR = 4.44), and solid component contact pleura (OR = 6.23) were independent risk factors for direct contact type. The model based on CT features demonstrated acceptable diagnostic efficacy in predicting VPI. The types of tumor-pleural contact are different, and there are differences in CT predictive variables.

问 AI

Bookmark

View Full Paper