# A Clinical–Radiomics Nomogram for the Preoperative Prediction of Aggressive Micropapillary and a Solid Pattern in Lung Adenocarcinoma

## TL;DR

This study creates a model combining clinical and radiomics data to better predict aggressive patterns in lung cancer before surgery.

## Contribution

A novel clinical–radiomics nomogram is developed for preoperative prediction of aggressive micropapillary and solid patterns in lung adenocarcinoma.

## Key findings

- The radiomics model outperformed the clinical model in predicting aggressive patterns.
- The combined clinical–radiomics model achieved the highest diagnostic accuracy with AUCs of 0.9186 and 0.9396.
- The comprehensive model showed enhanced clinical utility for guiding treatment decisions.

## Abstract

Background: A micropapillary pattern (MP) and solid pattern (SP) in lung adenocarcinoma (LUAD), a major subtype of non-small-cell lung cancer (NSCLC), are associated with a poor prognosis and necessitate accurate preoperative identification. This study aimed to develop and validate a predictive model combining clinical and radiomics features for differentiating a high-risk MP/SP in LUAD. Methods: This retrospective study analyzed 180 surgically confirmed NSCLC patients (Stages I–IIIA), randomly divided into training (70%, n = 126) and validation (30%, n = 54) cohorts. Three prediction models were constructed: (1) a clinical model based on independent clinical and CT morphological features (e.g., nodule size, lobulation, spiculation, pleural indentation, and vascular abnormalities), (2) a radiomics model utilizing LASSO-selected features extracted using 3D Slicer, and (3) a comprehensive model integrating both clinical and radiomics data. Results: The clinical model yielded AUCs of 0.7975 (training) and 0.8462 (validation). The radiomics model showed superior performance with AUCs of 0.8896 and 0.8901, respectively. The comprehensive model achieved the highest diagnostic accuracy, with training and validation AUCs of 0.9186 and 0.9396, respectively (DeLong test, p < 0.05). Decision curve analysis demonstrated the enhanced clinical utility of the combined approach. Conclusions: Integrating clinical and radiomics features significantly improves the preoperative identification of aggressive NSCLC patterns. The comprehensive model offers a promising tool for guiding surgical and adjuvant therapy decisions.