Movement correction by object recognition-based anatomical tracking in functional magnetic resonance urography (fMRU) - proof of principle

Holmberg, Andreas

Breathing motion is a challenge to analysis of imaging time series in many settings, especially in thorax and abdomen. We investigated in a software phantom as proof of principle, whether object recognition based tracking is capable of intensity-time-curve analysis. Images-time-series (no respiratory gating) of 100 kidneys were artificially generated (Matlab, TheMathworks, Natick, NA, USA). Respiratory movement was implemented by a sinusoidal coordinate shift with an amplitude of 3 cm and frequency of about 6 min-1. Renal intensity changes after contrast application were modeled using gamma functions for three anatomical compartments: cortex, pyramids and renal pelvis. Movement correction was carried out for half of the study population. Intensity-time-curves were extracted using automatically placed regions of interest relative to central coordinates of the kidney on the first image. Intensity changes over time extracted from the ROIs were subtracted from known intensity changes of the software model. Differences were assessed using Wilcox-Signed-Rank test for 50 kidneys with and 50 without movement correction. We used Bonferroni method to correct for multiple testing. Mean sum of differences between predicted and observed intensities across all kidneys and compartments was 0,072 with and 7,3 without movement correction. Significant difference between observation and model was not seen in any compartments of the tracking group (mean z-score: -0.8), whereas there was in 66 compartments in the non-tracking group (mean z-score: -3.2), signifying good agreement between theoretical model and observed intensity changes with object recognition-based tracking, and suboptimal agreement in the non-tracking-group due to movement artifacts. We conclude that object-recognition based anatomical tracking is feasible in fMRU as an alternative or addition to respiration gating. This may allow a higher temporal resolution of these studies in the future.

Cite

Citation style:

Holmberg, Andreas: Movement correction by object recognition-based anatomical tracking in functional magnetic resonance urography (fMRU) - proof of principle. 2018.

Rights

Use and reproduction:
No CC License (german copyright law applies)

Export