Survival Prediction of Locally Advanced Non-Small Cell Lung Cancer Via Machine Learning Techniques Based On Multi-Region Radiomics Features and Clinical Risk Factors
R Nishioka1, D Kawahara2, N Imano2, Y Nagata1,3(1)Medical and Dental Sciences Course, Graduate School of Biomedical & Health Sciences, Hiroshima University, Hirohima, JP(2)Department of Radiation Oncology, Graduate School of Biomedical Health Sciences, Hiroshima University Hirohima, JP(3)Hiroshima High-Precision Radiotherapy Cancer Center Hirohima, JP
Presentations
MO-IePD-TRACK 6-5 (Monday, 7/26/2021) 12:30 PM - 1:00 PM [Eastern Time (GMT-4)]
Purpose: To develop and validate a nomogram for predicting overall survival(OS) in patients with locally advanced non-small cell lung cancer(NSCLC) using multi-region radiomics analysis that analyze CT image in the image-based segmentation and the dosimetric-based segmentation.
Methods: From 2008 to 2018, 77 patients with locally advanced non-small cell lung cancer were included. A total of 66196 features were extracted from pretreatment computed tomography (CT) images. Rad-score was generated using Cox proportional hazards regression. The Clinical model was established by the use of clinical predictors which was selected by multivariable Cox regression model. The Nomogram model, combined with Rad-score and clinical predictors significantly associated with OS, compared with Rad-score and the Clinical model. The prediction model was evaluated by under the receiver operating characteristics curve, concordance index (C-index), calibration. Kaplan-Meier survival curve was calculated for OS comparing high-risk and low-risk cohorts based on nomogram scores.
Results: Thirteen features were selected to construct Rad-score, significantly associated with OS.Univariable and Multivariate analysis showed that Rad-score and Histologic type were independent predictors of OS, with p values of <0.001 and 0.01, respectively. Nomogram model showed best performance (C-index: 0.87 [95%CI: 0.81-0.93], AUC: 0.88) compared with the Clinical model (C-index:0.76 [95%CI: 0.64-0.88], AUC: 0.63) and Rad-score (C-index:0.84 [95%CI: 0.77-0.91], AUC: 0.88). Additionally, participants could be classified into low- and high-risk groups by the nomogram. The calibration plots also showed good consistency between the prediction and the observation.
Conclusion: A nomogram model, incorporate Rad-score and clinical predictors, is a useful tool for predicting OS in NSCLC patients which might facilitate clinical decision-making process.
ePosters
Keywords
Radiation Therapy, Radiation Effects, Lung
Taxonomy
TH- Dataset Analysis/Biomathematics: Machine learning techniques
Contact Email