USING ELECTRICAL STIMULATION TO RESTORE FUNCTION TO PARALYZED MUSCLES William Durfee Department of Mechanical Engineering University of Minnesota Kurt Korkowski, Ben Dunn, Karl Oberjohn, Brent Harrold Department of Mechanical Engineering University of Minnesota Michael Goldfarb, Heather Beck Karen Palmer, Jeff Chiou Department of Mechanical Engineering Massachusetts Institute of Technology Gary Goldish, Rich Scarlotto Physical Med. & Rehab. Service VA Medical Center, Minneapolis, MN Allen Wiegner, Nancy Walsh Spinal Cord Injury Service VA Medical Center, West Roxbury, MA HUMAN/MACHINE DESIGN LAB Department of Mechanical Engineering University of Minnesota (www.me.umn.edu/divisions/design/hmd/) • Stimulated Muscles = Power • Brace = Trajectory guidance u F • Brake = Control, stability T x,v Haptic interfaces for virtual product prototyping Active Element u IRC Activation Dynamics (2nd order) X CE Force-Length V CE Force-Velocity X Force Fscale • Passive Element X PE Force-Length V PE Force-Velocity • Smart orthotics + electrical stimulation for gait restoration Muscle mechanics OUTLINE What is Functional Electrical Stimulation (FES)? How FES can be used to restore motion State-of-the-art – Why it’s hard – Commercial products FES research at the U: – FES + "smart" orthosis – Modeling and control FES APPLICATIONS Bladder stimulation (incontinence) Cerebellar stimulation (movement disorders) Sensory substitution Visual prostheses (blindness) Auditory prostheses (deafness) Pain suppression (TNS) Pacemakers Limb control (paralysis) HOW FES WORKS Brain Spinal Cord Stimulator Limb SPINAL CORD INJURY NUMBERS 150,000 in U.S. 8,000 new cases each year 1/2 quadriplegic, 1/2 paraplegic LEADING CAUSES – – – – Automobiles Guns Sports (diving) Falls AGE GROUP Range: 15-29 Mean: 23 SPINAL CORD INJURY NUMBERS 150,000 in U.S. 8,000 new cases each year 1/2 quadriplegic, 1/2 paraplegic LEADING CAUSES – – – – Automobiles Guns Sports (diving) Falls AGE GROUP Range: 15-29 Mean: 23 FES BICYCLE ERGOMETER FreeHand FreeHand FreeHand (NeuroControl) Handmaster (NESS) Parastep (Sigmedics) WalkAide (Neuromotion) This is the story…...of the multimillion-dollar functional electrical stimulation (FES) project that got them walking….. It's about an ambitious but flawed technology and of questionable medical ethics. And it's about a tightknit research community so convinced of its promise that it would tolerate lessthan-acceptable standards of care for its human subjects. Inside Khawam's legs is a virtual birdnest of corroding electrodes ….. He and his doctors see a clear link between the infections and the electrodes….."Mr. Khawam's prognosis is decidedly poor…..either above-or belowthe-knee amputation….." New Mobility, June 1997 Wired For Walking: BY Sam Maddox Fifteen years ago, Bassam "Sam" Khawam, a 22-year-old Lebanese American living in the Cleveland suburbs, was paralyzed at T8-9 by a bullet. Khawam was big and physical, a black belt in karate. He was given the usual spinal cord injury prognosis: Get used to the chair, son, you're not going far without it. Khawam took the news the way many young, strong guys who join the gimp world do. He didn't buy it. "You are 22 and it happens to you," says Khawam, now a rehab engineer and father of two in Spokane, Wash. "You would want to walk again too." This is the story of Khawam and of a handful of other paralyzed research subjects, of the multimillion-dollar functional electrical stimulation (FES) project that got them walking, and of the scientist who lined up the money and ran the lab. It's a story of good intentions and good press on the side of science, and bad luck and bad faith as seen by the project's participants. It's about an ambitious but flawed technology and of questionable medical ethics. And it's about a tightknit research community so convinced of its promise that it would tolerate less-than-acceptable standards of care for its human subjects. Inside Khawam's legs is a virtual birdnest of corroding electrodes that cannot be removed without destructive surgery. He has had infections requiring antibiotics, plastic surgery and hospitalization. He and his doctors see a clear link between the infections and the electrodes, and one doctor suggested removing all of Khawam's thigh and calf muscle as the only way to get the hardware out. Another offers this bleak statement: "Mr. Khawam's prognosis is decidedly poor, as the future course of medical treatment may include either above-or below-the-knee amputation to rid him of a constant source of infection." LOWER LIMB FES FEXT ERNAL Inputs CONT ROL STIMULAT OR Measurements (The UNH Robot Lab, www.ece.unh.edu/robots/rbt_home.htm) IS IT LIKE A BIPED ROBOT? ROBOT CONTROL Commands CONTROLLER MOTORS SENSORS Desired Task SEGMENT DYNAMICS/KINEMATICS Disturbances FES CONTROL Cognitive feedback ? CNS PROCESSOR BRAIN VISUAL VESTIBULAR Commands ? SPINAL NATURAL SENSORS UPPER LIMB SEGMENT DYNAMICS/KINEMATICS MUSCLES Disturbances Spinal Lesion CONTROLLER NATURAL SENSORS ARTIFICIAL SENSORS Desired Task SPINAL CIRCUITS MUSCLES LOWER LIMB SEGMENT DYNAMICS/KINEMATICS Disturbances SERVOMOTORS AS ACTUATORS Torque Torque Current Speed Linear, time-invariant MUSCLES AS ACTUATORS Force = f(neural input, length, velocity, time, ...) F F Activation F Velocity Tim e High power/weight, but nonlinear, time-varying and uni-directional WHAT MAKES FES DIFFERENT? Not enough muscles Not enough sensors Muscle force too low Muscle fatigue Spasticity Weight constraints Size constraints Cosmetic constraints Ease-of-use constraints Reliability Implanted systems No sensory feedback User control? WHY MODEL? Inputs CONTROL STIMULATOR Measurements Complex system – Multi-link inverted pendulum – Nonlinear, time-varying actuators (muscles) Better models Better control Use model for: – Designing "generic" controllers – Prescribing/tuning custom systems MODEL VERIFICATION Direct comparison with experimental data Predictive capability Parameterization to the subject MODELING FOR CONTROL OF GAIT •Rigid body links (10) •Ideal joints •Passive torques •Active torques (muscles) MODELING MUSCLE u F x,v u T Active (AE) KSE CE KP BP “Muscle” = activity from single stim channel Joint-space model --> no knowledge of anatomy needed Passive (PE) 3 inputs (u,x,v), 1 output, modified Hill-type model ISOMETRIC MUSCLE Hammerstein model force stim Static nonlinearity Linear dynamic system Identify LDS with impluse response Identify SL by deconvolution MODELING MUSCLE Active Element u IRC u F Activation Dynamics (2nd order) x,v X CE Force-Length V CE Force-Velocity X S Fscale Passive Element X PE Force-Length S V PE Force-Velocity Force WHAT'S WRONG WITH THE MUSCLE MODEL Invariant F-A, F-L, F-V (no change with activation) Invariant twitch dynamics (uniform fiber types) Time-invariant (no fatigue) Zero neural time-delay Rigid SEC XMT XCE XSE CE KSE ISOLATED, ANIMAL MODEL MUSCLE 35 Model 30 Experiment Force (N) 25 20 15 10 5 0 0 4 8 12 T ime (s) (Durfee and Palmer, IEEE T rans. Biomed. Eng., 41(3):205-216, 1994) 16 INTACT, HUMAN MUSCLE (Abushanab: Ph.D. Thesis, MIT, 1995) EXPT VS. MODEL Knee flexion (deg) Hip flexion (deg) 40 30 20 0 2 4 Time (sec) Experiment Simulation 6 8 40 20 0 0 2 4 Time (sec) 6 8 WHERE WE ARE WITH MODELING & IDENTIFICATION Goal of modeling: simulation matches experiment Subject-to-subject variation is large ==> calibration is required How good is "good enough" will be determined by control strategy Must extend to subjects with SCI Better experimental ID methods evolving More diverse verification tests evolving PROBLEMS WITH FES-AIDED GAIT (1) Requires precise, stable control for repeatable steps Muscles are nonlinear, timevarying (2) Need to walk for reasonable distances Muscles fatigue rapidly PROPOSED SOLUTION: Stimulation plus "smart" orthotics BRACE (CBO) + FES • Stimulated Muscles = Power • Brace = Trajectory guidance • Brake = Control, stability CBO OVERVIEW, SPECIFICATIONS JOINTS – – – – 2-dof hip, 1-dof knee, fixed ankle hip adduction stop magnetic particle brakes Evoloid gear, 16:1 transmission STRUCTURE – aluminum, chromoly WEIGHT, INERTIA – 12.5 lbs, 10% of limb inertia STIMULATION AND CONTROL – 4-channel stimulation – on/off stimulation control – closed-loop brake control Designed for RESEARCH use (Goldfarb and Durfee, IEEE Tran Rehab Eng, 4(1):13-24, 1996) CBO EVALUATION PROTOCOL 4-channel stimulation (quad + peroneal) Parallel bars, walker 5 - 10 m lengths Compare gait with and without CBO Speed/distance, quadriceps use, repeatability Four subjects with paraplegia FES FES + CBO INCREASED SPEED, DISTANCE Gait Speed Gait Distance 0.14 60 50 0.09 40 Distance (m) 0.1 0.08 Speed (m/s) 50 0.12 0.12 0.06 0.04 30 20 0.02 10 0 0 Without CBO With CBO 25 Without CBO With CBO BETTER REPEATABILITY With CBO 120 120 100 100 Knee angle (deg) Knee angle (deg) Without CBO 80 60 40 20 0 0 0.2 0.4 0.6 0.8 1 1.2 Time (sec) 1.4 1.6 1.8 2 80 60 40 20 0 0 0.2 0.4 0.6 0.8 1 1.2 Time (sec) 1.4 1.6 1.8 2 OPEN ISSUES Substantial improvement in FES-aided gait.....but preliminary, laboratory results only Consumer-driven design (size, weight, ease of use) Technical issues Handling upper-limb inputs ??? Startle, stumble response ??? Fault tolerant equipment ??? Commercialization issues Market size ??? User acceptance ??? Who pays !!??