Association of radiomics and outcome in esophageal cancer following neoadjuvant chemo-radiation.

Abstract

146 Background: Radiomic data extraction from medical images has the potential to differentiate tumor phenotypes. We investigated the potential of esophageal cancer radiomic features to differentiate clinical outcomes following neoadjuvant chemoradiation (NACRT). Methods: We retrospectively reviewed 68 biopsy-proven mid-distal esophageal/gastroesophageal junction (GEJ) cancer patients treated from 2007-2014 with NACRT followed by esophagectomy on an IRB approved study in which patients were eligible if imaging was performed at our institution. The primary tumor was segmented on baseline computed tomography (CT) and 18-Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) scans with contour review by two independent radiation oncologists. A total of 126 quantitative 3-D shape and intensity-based radiomics features were extracted using in-house data algorithm software. Locoregional control (LRC), metastasis free survival (DFS), metastasis free survival (MFS) and overall survival (OS) were estimated by Kaplan-Meier analysis and multivariate Cox proportion hazard. Results: Estimated actuarial 5-year RFS, LRC, MFS, OS was 56.6% (mean: 67.8 months, 95% CI 53.7-81.8 months), 83.1% (mean: 92.8 months, 95% CI 81.4-104.2 months), 64.9% (mean: 76.4 months, 95% CI 62.2-90.7 months), 39.5% (median survival 40 months, 95% CI 24.59-55.40), respectively. Pathological complete response (pCR) was achieved in 34 (50%) patients. Tumor Regression Grade (TRG) 0-1 indicative of complete response was associated with improved LRC [p = 0.04; Hazard Ratio (HR) 4.24 (1.01-19.01)]. On univariate analysis, 16 PET-based and 14 CT-based radiomics features were found to independently predict pCR. On multivariate analysis (MVA), three PET-CT based radiomic features [sphericity (p = 0.04), compactness (p = .05) and PET_TGV (p = 0.08)] were found to closely predict pCR. When grouped by median, PET sphericity independently predicted RFS [p = 0.014; Hazard Ratio (HR) 0.31 95% Confidence Interval (CI) 0.11-0.83]. Conclusions: This study suggests pre-treatment radiomic features in esophageal cancer may be prognostic for response assessment (pCR) and overall outcome following NACRT.