Reports

Final Poster:

Executive Summary:

Medical conditions can be diagnosed in many different ways. Ear-related conditions are often first identified with a visual inspection of the middle ear canal and tympanic membrane. To do this examination, a doctor often uses an otoscope, a tool that magnifies and illuminates the ear canal, to conduct an examination, known as otoscopy, to quickly characterize the health of the patient’s inner ear. A common obstacle in teaching students how to conduct this simple visual inspection comes with a lack of understanding of what the inner ear looks like through an otoscope, as well as a lack of physical feedback knowledge, resulting in the student either applying too much pressure to the patient’s ear, or simply being too afraid of inserting the tool’s tip, the speculum, to an effective depth.

The novel otoscopy trainer proposed in this study will assist in a medical instructor’s ability to convey the principles required for conducting a more effective otoscopy exam. This preliminary analysis tackles the viability of using magnet-based systems to track the 6DOF state of the speculum tip and provide both the student and instructor with positional feedback, assisting in the students understanding of how motions of the otoscope handle translate to changes in the position of the otoscope head. Using live, magnetic localization data and orientation data from an inertial measurement unit (IMU), the system reconstructs the position and orientation of the tool in real time, providing instantaneous feedback using a simulated model tracking the otoscope user’s motions.

The trainer reconstructs the position of the otoscope using a system of three 3-axis magnetometers to find the relative magnetic field strengths measured at each sensor. A model is used to relate each observed magnetic field strength to a corresponding distance between the sensor and magnetic source.

||Hi||^2=K[(x-xi)^2+(y-yi)^2+(z-zi)^2]^(-3)*(3*((z-zi)^2+(x-xi)^2+(y-yi)^2+(z-zi)^2)+1)

With distance measurements from each 3D magnetic sensor, or magnetometer, the three distances are used to solve a system of three of the equations above to find the position of the point in 3D space (x,y,z).

Rigidly embedding a high-precision IMU into the handle of the otoscope allows for the real-time tracking of the orientation of the otoscope. The IMU outputs readings from a rate gyroscope, accelerometer and compass sensor. Using a complementary filter such as the Mahony filter, the 3D orientation can be reconstructed in the form of Euler angles.

With the x, y, z coordinates and Euler angles of the otoscope in 3D space, a system of transformation matrices is then used to apply the observed changes in position and orientation to a simulated replica of the otoscope in a 3D plot.

A simple test of the interpolation method was performed in this project. In this simple experiment, the magnetic flux densities from 3 magnet sensors are recorded as the otoscope is held constant at a 0^o position. Then, the otoscope is moved to a 10^o position clockwise from the original position, moving in only 1 DOF. The magnetic flux densities from each of the sensors are also recorded in this position. Because the otoscope is constrained in the z-direction, it is free to move in the x and y directions, as it rotates from 0^o to 10^o. From this data, the position at the 5^o angle is estimated through interpolation, generating a position error in the x and y directions of 0.63 mm and 2.95 mm. Therefore, this indicates that the goal of 0.5 mm accuracy is feasible due to the relatively high maximum position accuracy of 0.63 mm in 1 DOF through this test.

The efficacy of the position tracking was tested by moving the permanent magnet across a custom testing stand with known positions, resulting in an accuracy of 3.3 mm. Similarly, orientation tracking was tested by moving the IMU across a protractor and comparing the measured angle to the actual angle traveled by the sensor, resulting in an average orientation accuracy of 3.55 degrees. The targeted position accuracy of this model was 0.5 mm, however it was not achieved using the methods mentioned in this work. To improve the accuracy and reach this accuracy in a future product, a preliminary test of an interpolation method was carried out in this project and it demonstrated that the best possible accuracy was 0.63 mm in one degree of freedom, conveying that this model may be successful in achieving the targeted accuracy.

Presentations:

Individual Component Analysis Reports: