Radiation and Cancer Physics

SS 32 - Physics 10 - Imaging for Response Assessment II

231 - Pre-Treatment Magnetic Resonance Imaging Features Predict Post Chemo-Radiation Clinical Outcome for Patients With Local Advanced Rectal Cancer

Wednesday, October 24
7:45 AM - 7:55 AM
Location: Room 217 A/B

Pre-Treatment Magnetic Resonance Imaging Features Predict Post Chemo-Radiation Clinical Outcome for Patients With Local Advanced Rectal Cancer
X. Qi1, N. Li2, J. Jin2, and X. Pan3; 1Dept. of Radiation Oncology, UCLA, Los Angeles, CA, 2State Key Laboratory of Molecular Oncology and Department of Radiation Oncology, National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences (CAMS) and Peking Union Medical College (PUMC), Beijing, China, 3School of Computer Science and Technology, Xi'an University of Posts & Telecommunications, Xi'an, China

Purpose/Objective(s): Quantitative analysis of tumor image features may predict treatment outcome and guide individualized treatment. This work aims 1) to evaluate the performance of quantitative imaging features extracted from pre-treatment MRI, as well as clinical characteristics, for prediction of post-treatment response to chemo-radiation treatment (CRT) in locally advanced rectal cancer (LARC); and 2) to identify plausible pre-treatment prognostic biomarkers that are best associated with post-CRT clinical outcome.

Materials/Methods: Forty-five consecutive patients with LARC treated during 2015-2016 were included. Each patient received neo-adjuvant CRT with 50 Gy in 25 fractions, followed by total mesorectal excision surgery after completion of RT of 6-8 weeks. Other than planning CT, all patients underwent longitudinal pre-, during and post-RT MRI scans, including FRFSE T2 and diffusion weighted MRIs, etc. Image features were extracted from pre-treatment T2 images for the GTV, resulting in a total of 1838 quantitative features, along with clinical information (i.e., TNM staging etc.). According to clinical response assessed by post-operative pathology or MRI and colonoscopy, the patients were stratified into: (a) good responder group: pathological complete response (pCR) and partial response (PR); and (b) poor responder group: stable disease (SD) or progressive disease (PD). A novel nonlinear dimensionality reduction technique called T-distributed Stochastic Neighbor Embedding (T-SNE) was utilized to reduce high-dimensional tumor features, in which similar objects are modeled by nearby points and dissimilar objects are modeled by distant points, resulting a remaining of 182 features. Considering that the patient’s features are high dimensional and nonlinearly separable in the sample space, Gaussian kernel function was introduced to map the input space into a high-dimensional feature space. The features were then fed into the support vector machine model to create the predictive model in feature space. Five-fold cross validation method was utilized to train and validate the model. The model performance was evaluated using receiver operating characteristic (ROC) curve. We further identified the most important features (top 5%) that may affect prognosis.

Results: 54.7% of patients responded to the CRT: pCR (21.4%) or PR (33.3%), while 45.3% of patients achieved poor response, SD (33.4%) or PD (11.9%). Pre-treatment MRI features predicted post-CRT response with a calculated AUC = 0.82. Nine patient specific features, i.e., shape/roundness, morphology/GLCM texture, intensity, as well as clinical N stage, were identified as predominant predictors of response for LARC.

Conclusion: Pre-treatment image features, in combination of patient characteristics, could be used for post-treatment outcome prediction for LARC. Systematic analysis of image features via longitudinal multi-parametric MRIs should lead to improved predictive value to guide personalized treatment.

Author Disclosure: X. Qi: None. J. Jin: None.

Xiangrong Qi, PhD

University of California, Los Angeles

Disclosure:
Employment
Dept. of radiation Oncology, UCLA: Assist. Professor: Employee

Presentation(s):

Send Email for Xiangrong Qi


Assets

231 - Pre-Treatment Magnetic Resonance Imaging Features Predict Post Chemo-Radiation Clinical Outcome for Patients With Local Advanced Rectal Cancer



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 Pre-Treatment Magnetic Resonance Imaging Features Predict Post Chemo-Radiation Clinical Outcome for Patients With Local Advanced Rectal Cancer