3675 - Deep Learning-Driven Preoperative CT-Radiomics Model for Predicting Postoperative Radiation Therapy Benefit in the Elderly with NSCLC: A Multicenter Study
Presenter(s)
Z. Ma1, and Z. Hui2; 1Department of Radiation Oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China, 2Department of VIP Medical Services, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China
Purpose/Objective(s): The elderly make up the major proportion of both new cases and deaths of non-small cell lung cancer (NSCLC), yet they are often underrepresented in clinical studies. The effectiveness of postoperative radiotherapy (PORT) for the elderly with NSCLC remains unclear. Specific subgroups of these patients may benefit from PORT. We aimed to identify those most likely to gain from PORT.
Materials/Methods: Patients aged =60 with pN2 NSCLC after radical surgery between January 1, 2011, and June 30, 2019, were enrolled in one academic institution as the training set. Participants in a randomized controlled trial were included as the test set. Patients across another academic medical center were enrolled as external validation sets. Radiomics features were extracted from preoperative CT scans. The least absolute shrinkage and selection operator-Cox regularization model was used for data dimension reduction and feature selection. A radiomics index was developed using a deep learning model known as DeepSurv. The radiomics index's prediction capability was determined by the area under the curve (AUC) of the receiver operating characteristic. Overall survival (OS) was compared between patients who received PORT and those who did not, based on subgroups defined by the radiomics index.
Results: The training, testing, and external validation datasets comprised 512, 66, and 85 individuals. In the training cohort, the radiomics index effectively predicted OS, with an AUC of 0.82 (95% CI, 0.77-0.88) for the 5-year OS. The high-radiomics index group demonstrated significantly worse OS than the low-radiomics index group (5-year OS, 28.79% vs. 67.47%, p<0.01). Individuals were categorized into two risk groups using the median radiomics index. Patients in the high-risk group showed a significant benefit from PORT (5-year OS, 37.96% vs. 72.07%, p<0.01), while those in the low-risk group did not (5-year OS, 79.56% vs 82.75%, p = 0.25). The testing and external validation cohorts demonstrated consistent findings.
Conclusion: We developed a deep learning-driven preoperative CT-based radiomics index that can predict OS and assess the potential benefits of PORT for elderly patients with NSCLC, offering a personalized approach to PORT that optimizes outcomes while avoiding unnecessary treatment for low-risk patients.