2022 World Conference on Lung Cancer (ePosters)-EP13.01-002. Radiomic Signature on CT Images: A Noninvasive Biomarker for Pretreatment Discrimination of EGFR Mutations in NSCLC Patients

EP13.01-002. Radiomic Signature on CT Images: A Noninvasive Biomarker for Pretreatment Discrimination of EGFR Mutations in NSCLC Patients

Back to course

Pdf Summary

A study conducted by Kahya et al. aimed to evaluate the use of pretreatment computed tomography (CT) radiomics features as a noninvasive biomarker for predicting the Epidermal Growth Factor Receptor (EGFR) mutation status in patients with non-small cell lung cancer (NSCLC). The study included 430 patients who underwent CT examination and were tested for EGFR mutation. Radiomics models were constructed using machine learning methods and the performance was evaluated using receiver operating characteristic (ROC) curve analysis.<br /><br />A total of 1409 quantitative imaging features were extracted from the CT images. The Lasso algorithm was used for feature selection, and 89 features were selected. Three optimal features were identified using the Lasso algorithm. The XG Boost machine learning method had the highest area under the curve (AUC) for the training set (AUC 1, 95% confidence interval [CI] 0.91-0.98) and the test set (AUC 0.66, 95% CI 0.61-0.70). When the histopathological type parameter was added, the AUC values were 1 (95% CI 0.91-0.98) for the training set and 0.94 (95% CI 0.91-0.98) for the test set.<br /><br />The results suggest that radiomics modeling can predict the EGFR mutation status of tumors. This approach may be useful for patients who are not suitable for biopsy or when biopsy results are not adequate to detect EGFR mutations. In the future, radiomics may become an essential tool for determining the molecular subgroups of NSCLC.

Asset Subtitle

Yusuf Kahya

Meta Tag

Speaker Yusuf Kahya

Topic Pulmonology, Radiology, and Staging

Keywords

pretreatment computed tomography

radiomics features

noninvasive biomarker

EGFR mutation status

non-small cell lung cancer

machine learning methods

quantitative imaging features

Lasso algorithm

XG Boost machine learning

molecular subgroups