2913
Quantitative ADC as an early imaging biomarker of response to chemoradiation in esophageal cancer1The University of Texas MD Anderson Cancer Center, Houston, TX, United States
Synopsis
The goal of the study was to investigate if quantitative ADC can be used as an early imaging biomarker for predicting the treatment response to chemoradiation in esophageal cancer. Using pathological findings as the gold standard, our study demonstrated that the change in quantitative ADC from baseline (before treatment) to interim (two weeks after the initiation of treatment) was highly predictive of whether patients had residual tumors at the end of their treatment.
Introduction
Changes in diffusion-weighted images (DWI) and quantitative apparent diffusion coefficients (ADC) can provide a measure of the changes in the tissue microenvironment such as intracellular density and integrity of the extracellular matrix. Such changes often precede morphological changes and thus can be highly helpful for non-invasive and early assessment of the efficacy of cancer treatment [1]. Here, we report a preliminary analysis of an on-going clinical study, in which quantitative ADC changes in esophageal cancer measured before (baseline) and at two weeks after initiation (interim) of treatment by chemoradiation are investigated as an early imaging biomarker for treatment response.METHOD
Seventeen patients with stages IIa-IIIb esophageal cancer were included in this analysis of an on-going study. During a period of 5-6 weeks of standard chemoradiation, each study patient underwent three MRI scans at the baseline, interim, and end of their treatment, respectively. Patients included in this analysis underwent surgical resection after their last MRI. These patients were categorized into three distinct tumor regression groups (TRG) according to the pathological findings of their surgical specimen: TRG1 for 0% viable tumor cells (n = 4); TRG2 for <1% - 10% viable tumor cells (n = 8), and TRG3+ for >10% viable tumor cells (n = 5).
All MRI studies were performed on a GE 3.0 T whole body MRI scanner (MR750, GE Healthcare) with a 32-channel torso phased array coil. Each MRI study included two separate DWI scans: one with a conventional single short echo planar imaging (ssEPI) DWI sequence and another with a FOCUS DWI sequence. The typical scan parameters for ssEPI DWI were: b-values=0,200,800s/mm2, FOV=46x46cm, slice thickness=4mm, TR/TE=5000/70ms, matrix=96x128. The typical scan parameters for FOCUS DWI were: b-values=0,30,60,100,600s/mm2, FOV=16x12cm, slice thickness=4.5mm, TR/TE=4500/50ms, matrix=96x48.
The images by ssEPI DWI were fitted with a mono-exponential model to derive a quantitative ADC map (ADCmono). The images by FOCUS DWI were fitting with an IVIM model [2] to extract both perfusion and diffusion contributions to the DWI signals: , where f, and represent perfusion fraction, pseudo-diffusion coefficient, and true diffusion coefficient, respectively (3). Image analysis was performed using a software program developed in-house that has GUI-based features including image import, model fitting, semi-automated tumor contouring (using the snake or level-sets algorithms), and export of ADC statistics from the contoured regions. Using the software, tumors were contoured on the b=200 s/mm2 images for ssEPI DWI and on the b=100 s/mm2 images for FOCUS DWI.
Distributions of the relative changes (between the baseline and interim measurements) in contoured ADC parameters were compared for heterogeneity across the three TRG groups using a Kruskal-Wallis test. Inter-feature dependence among the relative changes was assessed using Spearman-rank correlation. Receiver operating characteristics (ROC’s) were further evaluated for the top summary ADC parameters to characterize the extent to which they could be used to discriminate responders (TRG1) from non-responders (TRG2+, defined as the combined TRG2/TRG3+ groups). A p-value of less than 0.05 was considered statistically significant.
RESULTS
When comparing the % change by the 3-factor TRG groups, the top 10 most statistically significant summary ADC parameters that were identified were found to be highly correlated. The heatmaps from the Spearman’s correlation matrix of the 10 ADC parameters for ssEPI DWI and FOCUS DWI are presented in Fig. 1a and 1b, respectively.
With a binary classification of patients as responders (TRG1) and non-responders (TRG2+), Youden’s Index indicated that the relative change in both volume mean ADC of the contoured regions by ssEPI DWI and volume ADC 90th percentile of the contoured regions by FOCUS DWI yielded perfect values of 1.0 for the area under the ROC curve (sensitivity =1.0 and specificity = 1.0). At a threshold of 27.7%, either of the two ADC parameters provided a complete separation between TRG1 and TRG2+.
DISCUSSION
Despite a small cohort, our study indicates that quantitative ADC can be highly useful as an early predictor of response to chemoradiation in esophageal cancer. Similar findings have been reported in a recent study of a comparable group of patients from a different institution [3]. From a clinical perspective, the ability to differentiate TRG1 from TRG2+ early on during treatment of esophageal cancer will be highly impactful as patients in TRG1 may forgo surgery and patients in TRG2+ may benefit from therapy intensification [4]. Future studies will help elucidate the inter-observer reproducibility of our findings.Acknowledgements
No acknowledgement found.References
[1]. Padhani AR, et al. Neoplasia 2009;11(2):102-125. [2]. Le Bihan D, et al. Radiology 1988;168(2):497-505. [3]. Van Rossum P, et al. Radiotherapy and Oncology 2015; 115(2):163-170. [4]. Lin S and Liao Z. Esophageal Cancer. Decision Making in Radiation Oncology. Springer Science and Business Media. 2010, Vol. 1; 329-58.Figures
