Automatic region tracking for MR glomerular filtration rate analysis

Contrast-enhanced dynamic Magnetic Resonance Imaging (MRI) acquisition is a common method to retrieve functional information from organs in the human body. Applied to the kidney, the observation of the signal evolution in the cortex of a MR-series gives access to the renal perfusion and filtration. The glomerular filtration rate (GFR) is in particular the most useful quantitative index of renal function. Since the rapid bolus passage hampers the use of gated sequences, fast sequences have to be employed to enable a data acquisition while free-breathing. As a result, the acquired data contains motion artifacts caused by the respiratory cycle, spontaneous movements and drifts which limit quantitative analysis of the data. Although these problems can in principle be addressed with motion correction algorithms applied in a post processing step, additional challenges arise from the fact that image amplitude changes not only due to motion but also due to the contrast change during bolus passage. This study proposes a 2D region tracking method for retrospective motion correction without sacrificing temporal resolution which addresses the latter point by a preparative learning phase.