# Prognosis from Pixels: A Vendor-Protocol-Specific CT-Radiomics Model for Predicting Recurrence in Resected Lung Adenocarcinoma

## TL;DR

This study shows that CT radiomics can accurately predict cancer recurrence in lung adenocarcinoma patients when using a standardized imaging protocol.

## Contribution

The paper introduces a reproducible CT-radiomics model specific to a uniform imaging protocol for predicting recurrence in lung cancer.

## Key findings

- Five radiomic features significantly predicted recurrence, including Shape Sphericity and GLCM Autocorrelation.
- The model achieved high AUC values (0.96–0.99) and strong predictive metrics on test data.
- Results suggest protocol-specific radiomics can improve recurrence prediction in lung cancer patients.

## Abstract

Radiomics is often limited by poor reproducibility across scanners and acquisition settings, which can reduce predictive performance. In this study, we investigated whether a CT-radiomics signature derived from a standardized CT image acquisition during routine preoperative evaluation can predict recurrence after complete resection of stage I lung adenocarcinoma while controlling for imaging protocol. Our goal is twofold: (1) to provide a practical framework for developing reproducible radiomic signatures Within a uniform image acquisition protocol, and (2) to demonstrate a protocol-specific model that could help identify patients who may benefit from closer surveillance or consideration of additional therapy.

Background: Radiomics can provide quantitative descriptors of tumor phenotype, but translation is often limited by feature instability across scanners and protocols. We aimed to develop and internally validate a protocol-specific CT-radiomics model using preoperative imaging to predict 5-year recurrence in patients with stage I lung adenocarcinoma after complete surgical resection. Methods: The retrospective study included 270 patients with completely resected stage I lung adenocarcinoma from January 2010–December 2021, among whom 23 (8.5%) experienced recurrence within five years. Radiomic features were extracted from routine preoperative CT scans. After preprocessing to remove highly constant and highly correlated features, the Synthetic Minority Over-sampling Technique addressed class imbalance in the training set. Recursive Feature Elimination identified the most predictive radiomic features. An XGBoost classifier was trained using optimized hyperparameters identified through RandomizedSearchCV with cross-validation. Model performance was evaluated using the ROC curve and predictive metrics. Results: Five radiomic features differed significantly between recurrence groups (p = 0.007 to <0.001): Shape Sphericity, first-order 90Percentile, GLCM Autocorrelation, GLCM Cluster Shade, and GLDM Large Dependence Low Gray Level Emphasis. The radiomics model showed excellent discriminatory ability with AUC values of 0.99 (95% CI: 0.98–1.00), 0.97 (95% CI: 0.91–1.00), and 0.96 (95% CI: 0.85–1.00) on the training, validation, and test sets, respectively. On the test set, the model achieved sensitivity of 100% (95% CI: 51–100%), specificity of 94% (95% CI: 81–98%), PPV of 67% (95% CI: 30–90%), NPV of 100% (95% CI: 90–100%), and overall accuracy of 95% (95% CI: 83–99%). Conclusions: Under protocol-homogeneous imaging conditions, CT radiomics accurately predicted recurrence in patients with completely resected stage I lung adenocarcinoma. External multi-vendor validation is needed before broader deployment.