While MR guided interventions were initially proposed more than 20 years ago [1], some of their aspects still need improvement. Notably, possible tissue motion during the interventional procedure leads to the desire to perform simultaneous high speed tracking of the surgical instrument and imaging. As both standard MR-based tracking and MR imaging rely on the same sets of scanner gradients and transmit coils for signal excitation and position encoding, this is typically not possible- one can either track or image at one given time. Few alternatives exist that can accomplish this goal; among them, the use of optical sensor linkages to measure torsion and flexion of fiber optic cables embedded in the hollow needle were shown to be able to predict the position of the proximal end of the needle, once the distal end position is accurately known [2]. In this work, the first steps towards a novel approach for simultaneous tracking and imaging, based on an inertial measurement unit (IMU) and technology from plane/missile tracking, are presented. The MEMS-based components of an integrated IMU (accelerometers and gyroscopes) are magnetic field compatible, and have been occasionally used for different purposes in the MRI environment. Different than in missile tracking, however, where the accelerometers’ noise sensitivity and the gyroscopes’ drift are mitigated by precise position updates using global positioning system updates, it is intended for periodic, accurate position updates to be provided by MR tracking.

The expected geometry and components of a surgical instrument (with the IMU and the miniature MR tracking coils to be located at the tip of the surgical instrument) are displayed in Figure 1a. The expected temporal interlacing of tracking and imaging is showcased in Figure 1b; in between these reference tracking scans, position updates for the instrument will be provided by the IMU at higher than 100Hz frame rate.

Initial experiments were performed to understand the feasibility and the challenges of using IMU’s for position tracking in high magnetic fields. A box containing a small IMU sensor (Adafruit LSM9DS0) and its attached circuitry was attached to a MR-compatible position encoder (Figure 2). Periodic motions over a 9cm region were performed on one axis for 1 minute in the fringe field of a 3T scanner, in a background field of ~1T, while recording the IMU outputs at 166Hz. Calibration for the zero level offset, sensitivity and misalignment in the three sensitivity axes of the sensors were achieved using the algorithms described in [3].

The position obtained from IMU’s (whose readings were filtered using a Kalman filter), was reset periodically to the absolute truth provided by the position encoder (in a manner similar to what will be accomplished by MR tracking); it was then compared with the position obtained from the position encoders. For one axis motion, Figure 3 displays the measured acceleration (bottom), the estimated velocity (3rd row), the ground truth (green) and the IMU-estimated position (blue) for every 2 second locking (first row). The second row of this graph displays the position error (difference between the green and blue curves on the top row). The root mean square of the position error, while “tracking” every 0.5, 1 and 2 seconds, is 2.1, 3.8 and 11.2s respectively. Note that accurate tracking tends to be achieved during the time the sensor moves (sloped curves in Figure 3 top); it is mostly during the time that the sensors are stationary (plateaus in Figure 3 top) that error accumulation occurs.This work presents preliminary evidence, indicating that instrument tracking using an IMU and relatively infrequent position updates using MR tracking is feasible. A higher quality IMU, coupled with an improved algorithm for bias estimation is likely to result in rms position errors of less than 2mm, using MR tracking updates less frequent than 1/s.