Andrew Gravunder, Andrew Hoffmaster (The Catholic University of America)
Abstract
The objective of the senior design project was to design and fabricate a low cost, in-home, robotic controlled therapeutic system that opens and closes the right hand of a stroke patient. This system will also incorporate bilateral hand therapy that requires the patient to exercise their unimpaired hand while using it to control the movements of the impaired hand. The intended result of this design is to provide a product for those patients continuing to suffer from hand impairments after the initial six week training period. The price expectation of this system is to be under $400.00 USD, making it a viable low cost in-home therapeutic solution for in-home patients.
Background
Every year over 795,000 people suffer from a stroke. Of those 795,000, 140,000 will die making stroke the third leading cause of death in the United States [1]. Of the stroke survivors only 10 percent will recover completely, 25% will have minor impairments, 40% will have moderate impairments, and 10% will be severely impaired and need 24 hour care [2]. One quarter of people who suffer from a stroke are under age 65 and will need medical care for the rest of their lives. Stroke costs in the United States are currently at $43 billion. $23 Billion is spent on medical and therapy costs and the remainder is spent on losses due to lack of work and functionality in society. The average cost of a stroke is $35,000 for the first 90 days of care. Of all of the money spent on stroke medical expenses 16% is spent on rehabilitation [3]. Robotic therapy currently costs $9,977 as opposed to only $8,267 for intensive traditional non robotic therapy [4]. Robotic therapy is currently very expensive, large, and requires the patient to be at a rehab center with trained professionals assisting in the rehabilitation.
Despite high costs there are advantages to robotic controlled physical therapy. First robotic therapy can limit the strain incurred by both patients and therapists. Traditional physical therapy usually consists in one or more physical therapists manually supporting a patient and moving an affected extremity. Robotic controlled therapy increases safety of a few risks that are present in traditional physical therapy, such as reducing risks of falling and limiting excessive forces applied on a stroke patient. Another major advantage of robotic therapy is that it allows for increased session time and frequency, which can stimulate better “re-wiring” of the brain due to increased repetition [5].
Hermano Igo Krebs is a principal research scientist in MIT’s Department of Mechanical Engineering. He believes that in-home robotic therapeutic devices will increase repetition and functionality of impaired extremities. “What you have to do is make more of them (robotic therapeutic devices), and that will drive costs to a point where people can have them in their homes.” [6]
Upper extremity hemiparesis is common condition found in post stroke survivors. About 30% to 36% of all individuals with hemiparesis still have arm impairments 6 months after the stroke was sustained [7]. Many stroke therapy techniques such as constrained-induced movement therapy have demonstrated positive benefits, but the disadvantage is being limited to strict inclusion criteria that specifically target patients with mild impairments [8]. An alternative method of therapy is bilateral arm training which has shown positive training outcomes for stroke survivors who have mildly, moderately, and severe motor impairments [9]. A recent study was performed by Mary Ellen Stoykov, Gwyn N. Lewis and Daniel M. Corcos that compared bilateral and unilateral training for Upper Extremity Hemiparesis in stroke. The used two control groups made up of twelve patients to undergo unilateral arm therapy and twelve patients that would undergo bilateral arm therapy. Their results indicated significant improvements in short –term bilateral training compared to unilateral training with patients who have mild motor impairments, and superior outcomes with long-term bilateral training with subjects who were much more impaired [10]. The major limiting factor to this study was the small control group size that only consisted of twelve subjects per group. This is a recent study and one of very few that addressed the comparative efficacy of bilateral training versus unilateral training using similar treatment techniques.
Design Objectives
To begin the design process important design parameters were identified and defined.
Use-ability: The exoskeleton hand system should be very easy to use since this device is being targeted for in-home use. In the home, the user will not have direct communication with a therapist on how to operate the exoskeleton hand system.
Safety: The targeted consumers for the exoskeleton hand system are stroke survivors that have sustained mild, moderate, or severe hand impairments. This means that safety is off ultimate importance being that the users will wear this device on their pre-existing injured hand.
Mobility: The exoskeletion system is going to be designed for in-home usage, thus the system will have to be mobile or easy to carry around by the user. The device is to be worn by user, so it cannot be too heavy or overbearing on the user. Durability: The exoskeleton system is an in-home device, so it will have to endure unforseen stress that come with in-home usage.
Cost: In order for this device to be marketable for in-home use, it will have to be one of the most inexpensive hand rehabilitation devices of its kind. The goal is to not exceed a $400.00 USD price limit, which is reasonable compared to other commercial products that are available.
Simplicity: In order to achieve the targeted price range, the design of the exoskeleton system must achieve a high level of simplicity. Simple design with the use of simple components must be achieved.
Table I: Ranking of Design Parameters
Design Options
Figure 1: Design Option 1. Simple Exoskeleton Design
This design creates only MCP extension. Most stroke patients that sustain hand impairment have more trouble opening the hand than closing it. This simple exoskeleton design uses two braces to support four fingers. The first brace wraps around the palm right above the wrist to the beginning of the MCP joint, and the second brace wraps around the four fingers above the MCP joint. Four cables will extend from the brace to brace and meet to form one cable. One large motor will then create extension force by pulling the main cable.
Figure 2: Design Option 2. Ring Exoskeleton Design
This exoskeleton design moves five fingers independently of one another. The exoskeleton is controlled by ten motors that pull the cables running through rings. The above apparatus is only shows the top part of the exoskeleton. There are also cables that run underneath the fingers and through the palm. All cables attach to the end of the finger tips so when the motors either control flexion or extension of a finger, the DIP and PIP joints will close and open. There are also braces on each side of every finger with pins to control rigidity of the fingers allowing them to open and close properly.
Design Option 3. Glove Exoskeleton Design
The final design option incorporates MCP, PIP, and DIP joint movement in both flexion and extension. It utilizes a normal tight glove made out of a thin comfortable material. Plastic pieces attach to the glove over the space between the MCP and PIP joints, the space between the PIP and DIP joints, and the region from the PIP joint to the end of the finger. The plastic pieces will be placed on top of the finger and the bottom of the finger, but the bottom will lack a plastic piece between the PIP and DIP joints so that the user can still flex the finger without obstruction. The plastic piece will have grooves in them to hold wire going from the motor to the anchor point. A total of eight motors are used, one for flexion in each finger, and one for extension in each finger (the thumb is excluded). The anchor point from the extension motors will be the end of the finger tip. The anchor point for the flexion motors will be the top of the PIP joint. This means the cable will run from under the finger over the fingertip to the PIP joint. When the cable is in this position it will allow for movement of the MCP and PIP and DIP joints. This design is favorable because of the wide range of motion the patient will receive. This design was the one chosen to pursue for the completion of the project objectives.
Figure 4: Design of Controller Glove
The controller glove will be a left hand standard commercially available glove that will be elastic and provide a tight fit to the user. The controller glove will have flex sensors in each of the fingers to accurately determine the position of the finger. The flex sensors provide a resistance of 10K Ohms when flat and a variable resistance of up to 110K Ohms when fully flexed. A 5 V supply will be sent into a 10k Ohm resistor in series with the Flex sensor. This creates a simple voltage divider which the output between the two resistors will be the voltage sent into the micro-controller to determine the position of the finger. We will use this same set up for all four fingers. The position of the finger found by the microcontroller will determine the amount of flexion of extension that the motors provide to the exoskeleton.
Methodology
After choosing the third “glove” design, a few modifications were made to make the design simpler and to fit the budget constraints. The initial design included moving the fingers independently with separate motors. To accomplish this each motor would be $20, and to purchase eight of them would consume half the budget, leaving little funds to focus on the rest of the project. So instead of purchasing eight smaller motors, two giant ¼ scale servo motors were purchased. To attach them to the middle, index, and ring fingers, springs at a junction point where used to keep the cables at the individual fingers at the same tension.
It was decided to not provide movement for the pinky because it proved very difficult to incorporate. The pinky is not as important as the other fingers. It is very difficult to leave your pinky extended while flexing your other fingers. This is because the fingers share a tendon in the forearm and unless you practice and force your pinky to resist the movement of the other fingers the pinky will naturally follow the movements of the other three fingers. To make the plastic pieces that provide the rigidity of the design ¾” polypropylene tubing was used. Three materials were tested: polyethylene, polypropylene, and polycarbonate and made finger prototypes out of each material. Polyethylene provided the lowest coefficient of friction, but it was observed that the friction was so low that the tools would often slip making it difficult to machine. Polycarbonate was hard and susceptible to fracture which made it not an ideal material. Polypropylene was easy to machine, had a low coefficient of friction so the cables could slide easily in the grooves, and provided the stiffness required. To make our final finger supports a lathe was used to make the thickness of the tubing 0.15” with a tolerance of +- 0.05”.
Figure 5: AutoCAD model of Finger Joint Pieces
Figure 5: 3D model of Finger Joints Pieces on hand
The bottom middle piece between the DIP and PIP joints is missing because it is impossible to close the hand if that piece is there. After shaving the tube to the desired thickness a mill was used to make two 1/16” grooves along the long axis of the tube. The tube was then turned 180 degrees and the process repeated.
Each individual finger piece was sewed into the glove and Velcro straps were utilized to keep the cables in their respective grooves and also to keep the pieces tight on the patient’s finger. In order to prevent unwanted wrist movement it is necessary to have a wrist splint. The splint is also the attachment point for the motors. We modified the splint by putting a plastic piece to provide rigidity and mounted the motors to the plastic pieces with screws.
Figure 6: Motors attached to wrist splint
Electrical System:
The electrical design encompasses a flex sensor and an Arduino UNO development board. A 5V signal was applied to the flex sensor while in series with a 10KΩ resistor to ground. The signal was measured at the node between the flex sensor and the 10KΩ resistor. This analog signal gave a value between 540 and 740. The Arudino has a built in servo library that was utilized to control the servo motors. A position is sent to the servo motors and the control algorithms in the Arduino Servo library make sure the position matches the command. A custom shield for the micro controller was fabricated to compact the design and allow for all of the electronic to be concealed in a small box instead of on a breadboard. A flow chart of the electronics is shown below.
Figure 7: Electronic Flow Chart
The most important design parameter for this project was safety. The first measure of safety is that the servo motors can only spin 180 degrees so it is impossible to extend the fingers too far. The second measure is the software is programmed to stop at 175 and 5 degrees so that the motors are not pushed all the way to the end. If these measures fail then two kill switches exist. The first kill switch cuts the power to the motors so that the motors will release tension. The second kill switch cuts the power to the microcontroller. Either kill switch will disable the motors and stops the force applied to the patient. Finally, the motors used in the B.E.S.T device are not strong enough to break bone. The strength of the cable is rated for 30-lbs of force which means the cable will break before any serious injury is incurred.
Results
A. Torque Analysis (pilot data)
A detailed theoretical force analysis of the exoskeleton device starts with the torque of the motors. Our Hiltec HS-805BB servo motors provide 274 oz-in (1.94 N-m) of torque at 5V while consuming s 2.3 A. The equation F_cable=Tmax*Rlever relates the torque of the motor to the force on the cable. The servo motors have a 1.75” arm to increase the force and allow for more motion. The total force each servo motor produces is 481 oz. (133.4 N). The force on each finger in an ideal world would then be one third of this value (since only three fingers are moving) or 160.41 oz. (44.6 N). To calculate the amount of theoretical torque on each joint, the equation Torque = Ffinger (force of finger) * Rpully (radius of pulley) was used.
The torque on the DIP joint was found to be: 80.21 oz-in (0.566 N-m)
The torque on the PIP joint was 120.31 oz-in (0.850 N-m)
The torque on the MCP join was 160.41 oz-in (1.133 N-m)
B. Motion Analysis
The motion testing of the B.E.S.T device occurred at the National Rehabilitation Hospital under Dr. Lee’s, the project’s faculty advisor, supervision. The test setup included two markers in between each joint and two markers on the top of the metacarpals. Three trials of five tests were taken with the B.E.S.T glove on, and three more trials of five tests were taken with a control human hand. The patient was told to move the controlling hand at slow medium and fast speeds for the tests, and the control patient moved her hands at the same speeds. The NDI equipment accurately measure the spatial positioning of markers and records the data into an excel file. To find the angle the x, y, and z coordinates are turned into a vector, and the vector dot product is used to determine the angle of each joint. The results of the testing proved that the B.E.S.T device achieved similar MCP movements as the control patient, and actually in some cases provided more MCP flexion than a normal hand. The B.E.S.T device achieved moderate PIP movement with a peak angle of flexion at 25 degrees as compared to a control patient with a peak angle of PIP flexion of over 60 degrees. The DIP flexion of the B.E.S.T device was virtually unnoticeable and the peak angle of PIP flexion was less than 10 degrees. The lack of DIP flexion is most likely due to joint rigidity and a separation of the bottom cable from the palm of the hand causing a non optimal force vector. Proposed solutions to the problem are: less joint rigidity in the form of new smaller finger pieces which do not require Velcro straps, and a plastic palm piece to prevent the bottom cable from separating from the palm too much.
Figure 8: The B.E.S.T. Slow Movement Analysis
Figure 9: The B.E.S.T. Fast Movement Analysis
Figure 10: Control Test Fast Movement Analysis
Figure 11: Control Test Slow Movement Analysis
C. Other Analysis
The device was fully functional and worked to open and close the hand of a control subject. The B.E.S.T device did not meet all of the original specifications. The aim was for the robot to mimic human movements in the MCP, DIP, and PIP joints. B.E.S.T achieves full extension of all of the joints which is where most stroke patients have difficulties. B.E.S.T does not achieve full flexion of the DIP and PIP joints. The MCP joint closes fully, but the PIP and DIP joints do not have much flexion. Other devices on the market achieve PIP and DIP movement, but fail to have any MCP flexion. The B.E.S.T device has MCP flexion and in the future will have full PIP and DIP flexion. The device does have good safety features, however more are necessary. Several people tried the device on and verified it as comfortable. The device is not easy to put on due to a large number of parts and straps. Expense specs were met with the B.E.S.T device costing $401.28 in parts with some funds allotted for previous prototypes and excess parts. The design was initially for four fingers, the device currently only moves three fingers.
Finished Prototype
Figure 12: Final BEST Device
Future Modifications
The first major modification is to gain PIP and DIP flexion. A bracket mounted onto the wrist support is necessary to keep the cables on the underside of the hand near the palm, so the distance from the cable to the motor will not decrease. The next modification would be to make new thinner finger stabilizers that are better secured to the glove. The materials considered are polypropylene, aluminum, and steel. Elastic bands would be used in place of the Velcro straps to make the glove easier to put on and would form fit to the user’s hand. The glove would need to be zippered to allow a stroke patient to use the device. A current monitoring system is in the works to monitor the current going into the motors and if the current gets too high an electronic relay would kill the power to the motors. Current into the servo motors is proportional to the amount of torque produced. The wrist splint needs to be longer and the motors need to be attached higher on the forearm to allow for longer cables. To force the hand of a stroke patient open it is necessary to have much stronger motors. Mechanical stops need to be employed to prevent over extension of the fingers and possible harm to the patient.
Discussion/Significance
Technical Feasibility:
The final prototype includes many parts that can be ordered from outside resources such as the gloves, wrist stabilizer, motors, and the Arduino microcontroller. These are the main parts essential for the fabrication of the exoskeleton. The technical parts that will provide the greatest difficulty in the production process are the specific finger pieces that provide channels for cables. The disadvantage of manufacturing these finger pieces is that the variability of the human hand is great. These pieces consist of multiple different lengths even for a specific user. The current exoskeleton hand was fabricated to serve the means of single “control” user and its performance is decreased for any other user with a different sized hand. This means that the production process would have to have a means of accommodating for the customization of the finger pieces. This means that a large scale production of this product would be a hassle and extremely expensive, therefore increasing the price of the exoskeleton on the market. And being that this product was designed to target in-home users the increase in price may not be economical.
To help provide a practical solution to this problem, the exoskeleton finger pieces would be produced to accommodate the current sizing that gloves can be purchased at on the market. These sizes are different among men and women. Therefore, instead of customizing each glove to a particular patient, the exoskeleton would be produced to fit each of the sizes observed for both males and females. The tradeoff will occur in quality for price in order for the exoskeleton to be produced at a large scale.
Economic Feasibility and Market Potential:
Labor expenses for the design and fabrication of the current BEST prototype has been calculated to be a total of $4,468, given that the average hourly salary for an entry level engineer was estimated at $20.22 /hour . This value was determined by summing total hours allocated to different actives: Background Research, Presentation/Report, Design Process, Purchasing of Materials, and Fabrication of Prototype. The total amount of hours spent accumulated to 221 hrs.
The design and prototyping of the BEST device will be expensive given that it is its initial stage of development. The current BEST prototype was fabricated at a cost of $401.28 and that is with an excess of materials. The end BEST product is estimated to range from $700 – $1000. This estimate is dependent on the cost of stronger motors that need to be utilized and the variability of multiple potential materials for finger pieces. Compared to the NESS H200 Electro Stimulation Device available on the market at $6,200, the BEST device is a feasible low cost solution for potential in-home customers.
Clinical Utility:
One of the objectives for this product was to an in-home and low cost therapeutic system that will mainly be used for post rehabilitation therapy, but there is also much potential for the BEST device to be used in a clinical setting. The BEST device incorporates bilateral training in which a stroke patient would exercise both impaired and unimpaired hands simultaneously. This form of training would augment current unilateral training and constrained stroke therapy techniques. Another advantage of the BEST device is that the physical therapist or clinician would be able to manipulate the movement of the hand exoskeleton on the patient by wearing the controller glove themselves. As observed in the photograph below, the BEST device is an ideal device that can be used in several exercises such as coordination training, such as a clinician tossing a ball to a patient.
Figure 13: BEST Device catching a ball via Control Patient
Control Patient Catching a Ball
Conclusion
The BEST device provides a viable robotic stroke therapy system targeted for in-home users at a low cost. The BEST device effectively opens the hand of a “control” patient, while providing full MCP joint flexion. Future modifications will be made to accommodate full PIP and DIP joint flexion. It was fabricated using the allotted $400, making its low cost an economic advantage as compared to other stroke therapy devices currently on the market. One of the BEST device’s greatest advantages is its potential for a wide range of stroke therapy techniques and applications. A device like this can serve many clinical purpose as well as in-home purposes, such as assisting various home activities needed to be done to improve a person’s quality of life.
Acknowledgements
Without an excellent group effort a project like this could not be completed. We also want to thank Dr. Sang Wook Lee for being an excellent faculty advisor. His guidance and direction was also a necessary component to this design project.
References
[1]The Stroke Center http://www.strokecenter.org/patients/stats.htm
[2]The National Stroke Association
http://www.stroke.org/site/PageServer?pagename=REHABT
[3]NINDS http://www.ninds.nih.gov/disorders/stroke/detail_stroke.htm#160131105
[4] Dr. Hermano Igo Krebs, MIT, http://www.mit.edu/newsoffice/2010/stroke-therapy-0419.html
[5] “Robotic Gait-Assisted Therapy in Patients with Neurological Injury,” Sayes,PhD, Crug, J PT; Missouri Medicine, Volume 105, Number 2.
[6] http://www.mit.edu/newsoffice/2010/stroke-therapy-04l19.htm
[7] Kwakkel G, Kollen BJ, van der Grond J, Prevo AJ. Probability of regaining dexterity in the flaccid upper limb: impact of severity of paresis and time since onset in acute stroke. Stroke. 2003;34:2181-2186.
[8] Mary Ellen Stoykov, PhD, Gwyn N. Lewis, PhD, and Daniel M. Corcos, Phd. “Comparison of Bilateral and Unilateral Training for Upper Extremity Hemiparesis in Stroke” . Neurorehabilitation and Neural Repair. Volume 23 Number
[9]November/December2009 http://nnr.sagepub.com/content/23/9/945.full.pdf+html
[10]Ibid