Lung Cancer

PV QA 4 - Poster Viewing Q&A 4

TU_36_3672 - Application of Radiomics Signature Captured from Pretreatment CT to Predict Brain Metastases in Stage III/IV Non-small Cell Lung Cancer Patients with ALK Mutation

Tuesday, October 23
2:45 PM - 4:15 PM
Location: Innovation Hub, Exhibit Hall 3

Application of Radiomics Signature Captured from Pretreatment CT to Predict Brain Metastases in Stage III/IV Non-small Cell Lung Cancer Patients with ALK Mutation
X. Xu1, L. Huang1, J. Chen1, J. Wang2, J. Wen1, L. Xie1, D. Liu1, J. Zhang3, and M. Fan1; 1Fudan University Shanghai Cancer Center, Shanghai, China, 2Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai, China, 3Department of Radiation Oncology,Fudan University Shanghai Cancer Center, Shanghai, China

Purpose/Objective(s): The purpose of this study is to investigate whether the radiomics approach could predict brain metastases (BM) for Stage III/IV non-small cell lung cancer (NSCLC) patients with ALK mutation.

Materials/Methods: Patients in stage III/IV NSCLC with ALK mutation confirmed by pathology from 2012 to 2017 were enrolled retrospectively. All patients’ CT before treatment were collected and the gross tumor volume (GTV) was defined by two experienced radiologists. Patients were divided into two datasets, one for model training and another for model test. A test-retest in RIDER NSCLC dataset was performed to find 132 stable radiomics features whose ICC value was more than 0.9. Then the LASSO Cox regression and a leave-one-out cross-validation were conducted to find optimal features for logistic regression model,which was then conducted to evaluate the predictive value of radiomics feature(s) for BM. AUC was calculated to evaluate model performance. Cox regression model was used to describe correlations between feature(s) with freedom from brain metastases (FFBM), and log-rank test was implemented for stratified analysis.

Results: Totally 132 patients were included in this study, among which 27 patients had BM at baseline examination. The median follow-up time was 12.7 (IQR 4.5-26.6) months. In the train set we found one radiomics feature (W_GLCM_LH_Correlation) significantly correlated with BM (P=0.014, AUC=0.6736) and it also had favorable performance in the test set (AUC=0.6417). The predictive value of radiomics feature could be further improved when combined with T stage and N stage (train set: AUC=0.73, test set: AUC=0.70). Those 105 patients without brain metastases before treatment were divided by their overall stage into stage III (57) and stage IV (48) groups. The radiomics feature had moderate performance to predict brain metastases during or after treatment in two groups (Stage III: AUC=0.6799; Stage IV: AUC=0.6556),implying that the feature correlated with pretreatment BM also had stable predictive power in the follow-up observation. What’s more, all stage III patients could be developed into low risk or high risk according to cox regression model developed by combing radiomics feature and treatment characteristics (chemotherapy and radiotherapy). Log-rank p-value was 0.021.

Conclusion: We find one radiomics feature derived from pretreatment chest CT to be predictable for brain metastases in stage III/IV NSCLC patients with ALK mutation, which could be beneficial to risk stratification for such patients. Further investigation is needed.

Author Disclosure: X. Xu: None. L. Huang: None. J. Chen: None. J. Wen: None.

Send Email for Xinyan Xu


Assets

TU_36_3672 - Application of Radiomics Signature Captured from Pretreatment CT to Predict Brain Metastases in Stage III/IV Non-small Cell Lung Cancer Patients with ALK Mutation



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Application of Radiomics Signature Captured from Pretreatment CT to Predict Brain Metastases in Stage III/IV Non-small Cell Lung Cancer Patients with ALK Mutation