Analysis of DWI-MRI

All images were imported to an in-house-developed computer program (‘Siswin’ version 8; S. Ringgaard, Aarhus, Denmark) for analysis. Image quality was rated from 0 to 5, based on the discernibility of the inner and outer borders (e.g. the visual distinction of the kidney from the surrounding tissue and calyces), cortex, medulla and artefacts for both kidneys, excluding images (slices) with a rating of 0. Cysts were visually defined and masked before further data processing. From the DWI scans, the Siswin software generated an ADC map and ADC was then measured directly on the ADC map.

We marked each kidney separately using the 12-layer concentric objects (TLCO) method [34]. The TLCO method has primarily been evaluated in renal blood oxygen level-dependent MRI with low intra- and interobserver variability, as reported elsewhere [34, 35]. If the right or the left kidney was not analysable, data from that kidney were omitted. The three outermost layers from both kidneys represented cortex, whereas layers 8–10 from both kidneys represented medulla. In sensitivity analyses, we included layers 2–4 and layers 3–5 to define cortex. The ΔADC was calculated by subtracting medullary ADC from cortical ADC. Examples of DWI and ADC images with and without the TLCO regions of interest (ROIs) can be found in ESM Fig. 1.