TU-C-12A-03: Functional MRI of Esophageal Cancer: Repeatability and Inter-Observer Reproducibility of Geometrically Corrected ADC Maps




In preparation to treatment response monitoring with diffusion weighted imaging (DWI) of neoadjuvant chemoradiation of esophageal cancer, we determined the repeatability and inter-observer reproducibility of the apparent diffusion coefficient (ADC). Geometrical corrections were performed prior to analysis, to enable multi-parametric analysis on the same ROI.


Ten patients with biopsy-proven esophageal cancer underwent an MRI exam before, during and after neoadjuvant chemoradiotherapy. The scan protocol consisted of an anatomical T2w scan (TE/TR = 100/1938ms), a DWI scan (b = 0, 200, 800 s/mm2, voxel size 3.25×3.25mm) and a B0 map to monitor magnetic field distortions (dual echo, dual acquisition, TE1/TE2 = 4.6/9.2ms). DWI scans were obtained with a single shot echo-planar read-out (BW_phaseenc ∼ 30 Hz/pixel). Two acquisitions were performed with opposing phase-encoding gradient directions, which results in opposing geometrical distortions. Subsequently, both DWI scans were geometrically corrected using the B0 map, and ADC maps were calculated. For analysis of the inter-observer reproducibility, two clinicians delineated the tumor volume on the T2w image. The ROIs were transferred to the ADC map. Finally, the repeatability was determined by comparing the ADC within one ROI between the two DWI scans in an exam. The inter-observer reproducibility was determined by comparing the ADC values within the two ROIs.


The average geometrical distortion over all ROIs was 2.67mm (range: 0.02-17.66mm). The repeatability of geometrical corrected ADC maps was 10.0%, the inter-observer reproducibility 7.9%. Both percentages are 95%-confidence intervals (1.96x standard deviation of median ADC).


Geometrical correction was required as considerable distortions were observed. Corrected ADC maps of esophageal cancer are associated with a modest repeatability error and inter-observer reproducibility error, which are in the same range compared to previous reports in other malignancies. In future, we will incorporate repeatability and reproducibility data in MRI-based (multi-parametric) treatment response models of esophageal cancer.