Encrypted login | home

Program Information

A Support Tensor Machine Based Algorithm for Distant Failure Prediction in Lung SBRT

no image available
S Li

S Li* 1,2 , B Li 1 , N Yang 1 , Z Zhou 2 , H Hao 3 , M Folkert 2 , K Westover 2 , P Iyengar 2 , R Timmerman 2 , H Choy 2 , S Jiang 2 , J Wang 2 , 1. Shool of Biomedical Engineering, Southern Medical University, Guangzhou, 510515, China; 2. The University of Texas Southwestern Medical Center, Dallas, TX, 75235, USA; 3. School of Computer Science and Technology, Xidian University, Xi'an, 710071, China.


TU-H-FS4-4 (Tuesday, August 1, 2017) 4:30 PM - 6:00 PM Room: Four Seasons 4

Purpose: To develop a support tensor machine (STM) based model for predicting distant failure in early stage non-small cell lung cancer (NSCLC) treated with stereotactic body radiation therapy (SBRT) using pre-treatment PET and CT.

Methods: The patient cohort used in this study includes 48 early stage NSCLC patients treated with SBRT from 2006 to 2012. Twelve patients (25%) failed at distant sites. For each patient, primary tumors are segmented on both pre-treatment CT and PET. Three-dimensional (3D) tensors of tumor for each imaging modality are constructed and used as the input for a STM-based classifier. A STM iterative algorithm is used to train weight vectors for every mode of the tensor for the classifier. Different from conventional radiomics-based approaches that rely on handcrafted imaging features, the STM-based classifier uses 3D imaging as the input for the model. Thus it takes full advantage of the imaging information. The STM-based predictive algorithm is compared to support vector machine (SVM) based method where the vectorization of tumor image is used as the input. Two conventional radiomics approaches that use 29 handcrafted imaging features as the input are also used for the comparison: 1) SVM-based classifier coupled with sequential feature selection (SVM-SFS); and 2) Logistic regression coupled with SFS (LR-SFS).

Results: A 10-fold cross validation strategy is employed for the performance evaluation. The STM-based predictive algorithm achieves highest area under the receiver operating characteristic curve (AUC=0.85), accuracy (78.75%), sensitivity (80%) and specificity (78.33%) among four methods investigated in this study.

Conclusion: We have developed a STM-based predictive algorithm using 3D image with an intrinsic 3D tensor structure as the input to predict distant failure of early stage NSCLC treated with SBRT. This STM-based method outperforms SVM with vectorization of image as the input, LR-SFS and SVM-SFS with handcrafted features as the input.

Contact Email: