GOAL: A novel magnetic resonance (MR) compatible accelerometer for respiratory motion sensing (MARMOT) is developed as a surrogate of the vendors' pneumatic belts. We aim to model and correct respiratory motion for free-breathing thoracic-abdominal MR imaging and to simplify patient installation. METHODS: MR compatibility of MARMOT sensors was assessed in phantoms and its motion modeling/correction efficacy was demonstrated on 21 subjects at 3 T. Respiration was modeled and predicted from MARMOT sensors and pneumatic belts, based on real-time images and a regression method. The sensor accuracy was validated by comparing motion errors in the liver/kidney. Sensor data were also exploited as inputs for motion-compensated reconstruction of free-breathing cardiac cine MR images. Multiple and single sensor placement strategies were compared. RESULTS: The new sensor is compatible with the MR environment. The average motion modeling and prediction errors with MARMOT sensors and with pneumatic belts were comparable (liver and kidney) and were below 2 mm with all tested configurations (belts, multiple/single MARMOT sensor). Motion corrected cardiac cine images were of improved image quality, as assessed by an entropy metric (p < 10-6), with all tested configurations. Expert readings revealed multiple MARMOT sensors were the best (p < 0.03) and the single MARMOT sensor was similar to the belts (nonsignificant in two of the three readers). CONCLUSION: The proposed sensor can model and predict respiratory motion with sufficient accuracy to allow free-breathing MR imaging strategy. SIGNIFICANCE: It provides an alternative sensor solution for the respiratory motion problem during MR imaging and may improve the convenience of patient setup.
GOAL: A novel magnetic resonance (MR) compatible accelerometer for respiratory motion sensing (MARMOT) is developed as a surrogate of the vendors' pneumatic belts. We aim to model and correct respiratory motion for free-breathing thoracic-abdominal MR imaging and to simplify patient installation. METHODS: MR compatibility of MARMOT sensors was assessed in phantoms and its motion modeling/correction efficacy was demonstrated on 21 subjects at 3 T. Respiration was modeled and predicted from MARMOT sensors and pneumatic belts, based on real-time images and a regression method. The sensor accuracy was validated by comparing motion errors in the liver/kidney. Sensor data were also exploited as inputs for motion-compensated reconstruction of free-breathing cardiac cine MR images. Multiple and single sensor placement strategies were compared. RESULTS: The new sensor is compatible with the MR environment. The average motion modeling and prediction errors with MARMOT sensors and with pneumatic belts were comparable (liver and kidney) and were below 2 mm with all tested configurations (belts, multiple/single MARMOT sensor). Motion corrected cardiac cine images were of improved image quality, as assessed by an entropy metric (p < 10-6), with all tested configurations. Expert readings revealed multiple MARMOT sensors were the best (p < 0.03) and the single MARMOT sensor was similar to the belts (nonsignificant in two of the three readers). CONCLUSION: The proposed sensor can model and predict respiratory motion with sufficient accuracy to allow free-breathing MR imaging strategy. SIGNIFICANCE: It provides an alternative sensor solution for the respiratory motion problem during MR imaging and may improve the convenience of patient setup.