Quality Control of MRI-based Kidney Volume Estimation in ADPKD patient in clinical practice

Total kidney volume (TKV) is a critical biomarker for monitoring disease severity in autosomal dominant polycystic kidney disease (ADPKD). TKV is typically measured by manually contouring kidneys on one sequence from abdominal MRI without any quality control. Here we show that by using a deep learning model to measure kidney volume on 5 routinely acquired abdominal MRI sequences, it is possible to apply outlier analysis to find images with acquisition artifacts and to correct or exclude them from TKV estimation. This improves volume measurement consistency among the 5 sequences from 4.3% to 1.3% after quality control.