# A nomogram model integrating radiomics and clinical variables to predict napsin a expression in lung adenocarcinoma patients

## TL;DR

This study creates a non-invasive model combining imaging data and clinical factors to predict a key marker in lung cancer, improving diagnosis and treatment planning.

## Contribution

A novel nomogram integrating radiomics and clinical variables for non-invasive prediction of Napsin A expression in lung adenocarcinoma.

## Key findings

- The integrated model achieved high AUC values (0.844 in training, 0.845 in validation) for predicting Napsin A expression.
- Calibration curves confirmed strong agreement between predicted and observed outcomes.
- The model showed balanced accuracy of 82.1% and 80.6% in training and validation cohorts, respectively.

## Abstract

Lung adenocarcinoma, a major subtype of non-small cell lung cancer, requires non-invasive diagnostic tools to improve early detection and differentiate primary from metastatic tumors. Napsin A, a key marker for primary lung adenocarcinoma, is traditionally assessed via invasive biopsy, limiting its utility in reflecting tumor heterogeneity. Radiomics, which extracts quantitative features from medical images, offers potential for non-invasive prediction of molecular markers like Napsin A.

To develop and validate a nomogram integrating radiomic features and clinical variables for non-invasive prediction of Napsin A expression in lung adenocarcinoma.

This retrospective study enrolled 308 lung adenocarcinoma patients (training cohort: n = 246; validation cohort: n = 62), with contrast-enhanced CT images were used to extract 1,734 radiomic features, which underwent dimensionality reduction via t-tests, Pearson correlation, minimum redundancy maximum relevance (mRMR), and LASSO regression, retaining 27 final features; significant clinical variables (gender, smoking history, pulmonary cavity, spiculation sign, pleural indentation sign) were selected by logistic regression. A nomogram integrating radiomic and clinical predictors was developed and evaluated using ROC curves (AUC for Napsin A prediction), calibration curves (Hosmer-Lemeshow test), and decision curve analysis (DCA) for clinical utility.

The integrated nomogram model outperformed standalone radiomic and clinical models in predicting Napsin A expression, achieving AUC values of 0.844 (95% CI: 0.790–0.898) in the training cohort (n = 246) and 0.845 (95% CI: 0.724–0.967) in the validation cohort (n = 62), with balanced accuracy of 82.1% and 80.6%, respectively. Calibration curves showed strong agreement between predicted and observed outcomes (Hosmer-Lemeshow P > 0.05), and decision curve analysis confirmed its superior clinical utility across diverse threshold probabilities.

The integrated nomogram offers a reliable non-invasive method for predicting Napsin A expression in lung adenocarcinoma, supporting personalized treatment decisions and reducing reliance on invasive biopsies.