Abstract B10: Radiomics signatures on the region defined by using multi-window CT to improve detection lung cancer screening
Lists
Only you see this
Abstract
Abstract Purpose: To evaluate the utility of defining a region using complementary information provided by different CT windows on the lung nodules to predict malignancy. Methods: We obtained 50 cases from National Lung Screening Trial (NLST) with 23 malignant and 27 benign patients. We segmented lung nodules using lung and mediastinal window independently and obtained the difference region (3D). About 260 quantitative radiomics features were exacted on the voxels segmented from different windows and the difference region. We used support vector machines (SVM) to build a predictive model to identify malignancy. We compared the effectiveness of the multi-window based features against single-lung window features using a 2-fold cross validation. Results: We find several radiomics features in the difference region to be predictive of malignancy; the best univariate features was Run_Length_Nonuniformity differnce region had an AUC of 0.911.The best feature in the lung window was defined by Sum_Entropylung window, which had an AUC of only 0.825. While using cross-validataion (2-fold) we obtained a mean accuracy of 80.97% and an AUC of 0.8 (specificity and sensitivity of 0.887 and 0.712). Compared to single-window, the best predictive features had accuracy of 78.17% and AUC of 0.772 (specificity and sensitivity of 0.876 and 0.667). Conclusions:The multi-window based radiomics features, especially the features from difference area, can improve diagnostic performance in differentiation of benign and malignant lung nodules to great extent. Citation Format: Hong Lu, Wei Mu, Yoganand Balagurunathan, Jin Qi, Mathew Schabath, Robert James Gillies. Radiomics signatures on the region defined by using multi-window CT to improve detection lung cancer screening [abstract]. In: Proceedings of the Fifth AACR-IASLC International Joint Conference: Lung Cancer Translational Science from the Bench to the Clinic; Jan 8-11, 2018; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2018;24(17_Suppl):Abstract nr B10.
BibTeX
@inproceedings{lu2018abstract,
author = {Hong Lu and Wei Mu and Yoganand Balagurunathan and Jin Qi and M. Schabath and Robert J. Gillies},
title = {{Abstract B10: Radiomics signatures on the region defined by using multi-window CT to improve detection lung cancer screening}},
booktitle = {Clinical},
year = {2018},
doi = {10.1158/1557-3265.aacriaslc18-b10},
}Loading related content...