2024 World Conference on Lung Cancer (WCLC) - ePosters-EP.08F.13 Machine Learning Analysis of Cardiac Sub-Structure Dose and Overall Survival of NSCLC Pts Treated with Concurrent Definitive Chemoradiotherapy

EP.08F.13 Machine Learning Analysis of Cardiac Sub-Structure Dose and Overall Survival of NSCLC Pts Treated with Concurrent Definitive Chemoradiotherapy

Back to course

Pdf Summary

The document outlines a study conducted by Thomas J. Dilling and colleagues at Moffitt Cancer Center on the impact of cardiac sub-structure radiation doses on overall survival in non-small cell lung cancer (NSCLC) patients treated with concurrent chemoradiotherapy. The study involved 418 patients treated between 2009 and 2020, with some receiving durvalumab. The analysis aimed to understand the significance of doses to the heart and specific cardiac substructures, which has been a challenging area of research due to limited data availability.<br /><br />Cardiac substructures were auto-contoured using Platipy software, and various dose metrics were considered. The research utilized both univariate and multivariable analyses, incorporating machine learning methods like Elastic Net Regression, Random Survival Forest, Gradient Boosting, and Survival Kernel Support Vector Machine (SVM) to analyze the data.<br /><br />The main findings identified certain clinical features, like the delivered dose, age, receipt of durvalumab, and tumor characteristics, as significant predictors of overall survival. In terms of cardiac substructures, the minimum doses to the right coronary artery and AV node were noted as important factors in the 'winning' SVM model.<br /><br />The discussion suggests that while clinical features were predominant in three out of the four models used, the cardiac features were most significant in the Random Survival Forest model. The study acknowledged the complexity of the relationship between cardiac substructure doses and overall survival, indicating the need for future models to include data on pre-existing cardiac disease.<br /><br />Overall, the study advances the understanding of how radiation doses to specific cardiac substructures might influence survival in NSCLC patients, utilizing innovative machine learning techniques to parse complex datasets.

Asset Subtitle

Thomas Dilling

Meta Tag

Speaker Thomas Dilling

Topic Local-Regional Non-small Cell Lung Cancer

Keywords

cardiac sub-structure

radiation doses

overall survival

non-small cell lung cancer

NSCLC

concurrent chemoradiotherapy

machine learning

Elastic Net Regression

Random Survival Forest

Survival Kernel SVM