185-LB: Repeatability and Reproducibility of Pancreas Volume Measurements Using MRI

Pancreas volume, as measured by magnetic resonance imaging (MRI), is smaller in individuals with newly-diagnosed type 1 diabetes (T1D) and declines over the first year after diagnosis. However, the repeatability and reproducibility of these MRI measurements is not established, and for MRI to be utilized to measure pancreas volume, the variation between measurements of pancreas volume must be determined. To determine the test-retest repeatability of pancreas volume measurements by MRI, we performed two or three same-day repeat MRIs in the same individuals with or without T1D (n = 8, mean age 24 y, range 11-40 y). A single radiologist, blinded to whether the subject had T1D or not, outlined the pancreas on these repeat images. Pancreas volume was determined by multiplying the area of the pancreatic regions of interest by the MRI slice volume (5 mm). The repeatability of test-retest scans had an intraclass correlation coefficient (ICC) of 0.979 (95% CI 0.928 to 0.996). To determine the inter-reader reliability of pancreas volume measurements, the pancreas was independently outlined by two board-certified radiologists (n = 11). Comparing results from multiple radiologists yielded an ICC of 0.931 (95% CI 0.776 to 0.981), with greater differences in measurement in larger pancreases (p < 0.005). The mean DICE coefficient, a measurement of the degree overlap between pancreas regions of interest between the two readers, was 0.77. These data show that measurements of pancreas volume obtained using MRI are both repeatable and reproducible across multiple readers with ICCs corresponding to excellent clinical significance (ICC > 0.9). These results indicate that MRI measurements of T1D pancreatic volume can be reliably determined and could be useful for clinical investigation.