Nomogram integrating MRI radiomics and white matter hyperintensity grading for predicting overall survival in patients with non-small cell lung cancer and brain metastases receiving whole-brain radiotherapy.

OBJECTIVES

To develop and validate a nomogram integrating radiomic features, white matter hyperintensity (WMH) grading, and clinical factors for predicting overall survival (OS) in patients with non-small cell lung cancer (NSCLC) and brain metastases (BMs) receiving whole-brain radiotherapy (WBRT).

METHODS

One hundred and forty-nine patients with BMs were enrolled. A radiomic score (Rad-score) was developed based on features identified by univariate Cox regression and least absolute shrinkage and selection operator Cox modeling. WMH was graded using the Fazekas scale, into mild- and extensive-burden groups. Cox proportional hazard models incorporating different combinations of these features were developed and compared. A nomogram was constructed by integrating the Rad-score, WMH grade, and independent clinical variables, and its performance was evaluated using receiver operating characteristic (ROC) analysis, decision curve analysis (DCA), and calibration plots.

RESULTS

Twelve radiomic features were identified. Both the Rad-score (hazard ratio (HR) = 1.072; p < 0.001) and WMH grade (HR = 2.420; p < 0.001) independently and significantly predicted OS. Significant improvements in model fit followed the inclusion of these variables (p < 0.01). The integrated nomogram outperformed the single-feature models, yielding area under the curves (AUCs) of 0.820, 0.900, and 0.889 for 1-, 2-, and 3-year OS, respectively, with a concordance index (C-index) of 0.706 (95% CI: 0.598-0.815). Calibration plots and DCA further supported the predictive accuracy and clinical utility.

CONCLUSION

The radiomics- and WMH grading-based nomogram represents a potential prognostic tool for patients with NSCLC and BMs receiving WBRT.