DeepLung: A Novel Lung Cancer Recurrence Prediction Model Using Deep Learning.

Author(s): Kairui Yang

Lung cancer recurrence represents a critical determinant of patient prognosis, posing a significant threat to survival outcomes. The development of reliable recurrence prediction tools is therefore clinically imperative to guide therapeutic decision-making and improve both survival and quality of life. The model comprehensively analyzes tissue features from all designated regions of interest (ROIs) identified in pathological reports to predict lung cancer relapse probability.Validation through timedependent receiver operating characteristic (ROC) analysis demonstrated robust predictive performance. Survival analysis using semi-parametric Cox proportional hazards models confirmed the model's superiority over conventional TNM staging, with statistically significant improvements in AUC values (p<0.05). This prediction model exhibits substantial clinical translational potential, providing a valuable foundation for personalized treatment strategies and emerging as a novel decision-support tool for prognostic management.