MR-guided breast biopsies are lengthy, expensive procedures, whose accuracy and false negative rates are of concern [1]. One of the reasons for their complexity is the fact that they are performed without real-time guidance; the lesions can be visualized for ~10 minutes after the contrast agent is injected, while the woman is inside the MRI magnet. The biopsies are conducted, however, outside the MRI magnet. One or more image confirmation steps are usually performed, to ensure that the correct tissue is biopsied. The goal of the work presented here is to assess whether a combination of 3-axis accelerometers, gyroscopes and magnetic field sensors— can provide a basis for accurate and inexpensive tracking of a surgical instrument in the fringe field of a MRI scanner.To understand the range of field sensors needed and the relationship between magnetic field and position, magnetic field simulations were first performed using Comsol. Figure 1 displays simulated radial (Br) and longitudinal (Bz) fringe fields for a 3T magnet, between the exit of the bore (z=1m) and the end of the table (z=3m) as a function of radial (r) and longitudinal distance from isocenter (z); this data indicate that all field components will remain below 500G for at least 110cm from the end of the table, in the region where breast biopsies are performed. The setup used for our preliminary data collection is presented in Figure 2. A translation stage with position encoders was placed next to a breast coil, at an approximate location for breast biopsies, in the fringe field of a 3T, MR750 GE scanner. A 3-axis accelerometer/gyroscope (ST LSM9DS0), and three orthogonal, range-appropriate [0-500G] Hall-effect sensors (Microswitch SS94A1) were attached to the translation stage. The sensor recordings were converted into position by implementing a Kalman filter on the accelerometer/magnetometer data for the axis of motion, and assuming a linear relationship between magnetic field and position. Calibration for the zero level offset, sensitivity and misalignment in the three sensitivity axes of the sensors were achieved using the algorithms described in [2].The position given by the position encoder and the magnetic field reported by the sensor as a function of time (while moving the translation stage over ~8cm) are shown in Figure 3. This graph confirms the linear relationship between magnetic field and position for the small distances traveled in breast biopsies. The steep field gradients and the sensor sensitivity (5mV/G) confirm that mm tracking accuracy is likely achievable. Figure 4 top shows the predicted/measured position from the system of sensors for a 1D translation over 10cm/70seconds; the 2 curves completely overlap when using a display scale that covers the range of motion (Fig 4 top). The ground truth was provided by the position encoder; the predicted position was obtained as described in Methods. Figure 4 also shows the estimated error (middle) and velocity (bottom). For our experimental 83Hz sensor update rate, the position root mean square (rms) error was 1.3mm; due to noise associated with the inertial sensor, artificially reducing the Hall effect sensor update rate by a factor of 2 increased the rms error to 1.6 mm. Increasing the sampling rate of the aiding signal by a factor of 2—by, e.g., employing a faster A/D card-- is expected to reduce the rms position error to 1mm.This work presents preliminary evidence, indicating that high precision instrument tracking in the fringe field of a magnet, using a combination of accelerometers, gyroscopes and range appropriate Hall-effect sensors, is feasible. Extension of the current 1D position prediction algorithm to 3 dimensions, by incorporating the orientation of the sensors, will be pursued next.