Breathing Motion Compensation for Medical Robots
Project By: Jason Hall
Project Adviser: Dr. Eftychios Christoforou
The information included here is a quick introduction and overview to my project. For additional information, please read the full project write up, via the links at the bottom of the page.
Background
Robotics in the medical field are used for many different kinds of procedures and operations: biopsies to sample potential tumors, ablations to burn away tumors, minimally invasive surgeries, neurosurgery, hip replacements and many other kinds of operations.
There are two main ways of controlling these robots. One is autonomously computer-controlled and one is to use a Haptic interface, which is a model of the actual robot, acting like a joystick. The autonomous method is good for tasks that have high repeatability, that require high amounts of accuracy and repeatability, while the hand-controlled method is good for unique tasks.
When a person breathes, their stomach and chest move in and out, and their internal organs shift. These motions need to be accounted for to ensure that the operation is a success and does not injure other parts of the patient.
Magnetic Resonance Imaging (MRI) scanners enable doctors to see the tissues inside a patient's body. To accurately perform a biopsy or ablation on the tissue, it would be helpful to doctors to see where the needle is located in three-dimensions, using the MRI machine to guide the operation. Currently they use a ultra-sound imaging machine to help position the needle, but this technology is limiting.
Project Overview
 |
| Breathing Bellows Belt |
For this project, we want to develop a method to allow all robots in the medical field to adjust to the patient's breathing, regardless of where the robot is used. To make this feasible, we have to use the most restrictive requirements, which for our purpose is use inside MRI machines, where the tight space and the high magnetic fields are a restricting set of conditions.
The first step is to determine when a patient is breathing in and out. There are many different techniques to achieve this, such as using an air filled mattress, which changes pressure when a person's back moves from breathing. This method is not usable because it takes up too much space in the MRI machine, and it is not a stable platform for operations. Just imagine performing surgery on a water bed.
Another method is to use a piezo-electric belt, similar to what comes with many exercise machines, attached around a persons chest or stomach. When a person breathes, their chest expands, stretching the belt, which creates an electric charge. That charge can be read by the computer controlling the robot. This piece of hardware is not compatible to be used with MRI machines because it contains high amounts of metal.
The technique we chose is to use breathing bellows. This is a small accordion tube attached around a patient's chest, connected to a pressure sensor by a tube. The bellows are already used with MRI machines to trigger an image capture, so that images are all taken at the peak of a person's breath. By using this MRI compatible belt with a pressure sensor, we can trace when a person breaths and adjust the robot accordingly.
Experimental Setup
 |
| Time Delayed Image of Arm Movement |
For our project, we used a simple one rotational degree-of-freedom robotic arm, to test whether our project would be successful. The plan is to have the end of the arm move the desired distance.
The computer that is controlling the robot is running Matlab with Simulink, and is connected to the robot via a MultiQ-PCI controller card. This card has digital-to-analog and analog-to-digital converters, which allows us to utilize many different kinds of sensors.
Experimental Results
 |
| Filtered and Unfiltered Input Signal |
Reading the input from the pressure sensors resulted in very high frequency noise that rendered the signal unusable. A Butterworth filter was implemented to reduce the noise and was adjusted so that enough noise was removed, while lag introduced was kept to a minimum.
The signal from the pressure sensor is in volts and needs to be converted to the desired height change from the patient's stomach or chest. This is done with a quick calibration, recording the reading of the sensor and corresponding height of the stomach, for when the patient breathes out and in. The desired height can be found by interpolating between the high and low sensor readings.
Future Improvements
The technique used to calibrate the sensor to the height of the patient's stomach is not completely accurate, relying on a right-angle ruler against a wall, resting on the patient and marking the change in height on the wall. A more reliable method, such as using the MRI machine directly to measure the height, or a laser range-finder would lead to greater accuracy in the calibration.
A person's stomach does not only move in and out while breathing, it also expands outwards in all directions, while the organs can shift up and down inside a patient's body. Using the MRI images and a processor to track the displacement in real time would improve this technique many fold.
Links
Photographs
 |
 | |
| Breathing Bellows on a Patient | Breathing Bellows on a Patient | |
 |
 | |
| Pressure Sensor | Calibration Technique | |
 |
 | |
| Robotic Setup | Patient with Robot | |
 |
| |
| Butterworth Filter Comparison | Unfiltered and Filtered Input Signal | |
 |
 | |
| Desired and Actual Motor Position | Desired and Actual Motor Position Difference | |
Movies
| Robotic Arm Motion | Robot Maintaining Distance from a Patient |