bioeng486.md

typora-copy-images-to
./img

BIOENG486

GNU General Public License v3.0 licensed. Source available on github.com/zifeo/EPFL.

Fall 2017: Sensorimotor Neuroprosthetics

[TOC]

Motor control of reaching and grasping

• clinical goal : fist restore main actions

• avoid

  • biomimetic : mechanical structure different per species
  • plasticity : not an excuse for bad results

• upper limb : exploration, reaching

  • 3 bones
  • 3 joints
  • 7 degree of freedom
  • at least 46 muscles

• hand : sensorial, manipulation, grasping

  • 19 bones
  • 16 joints
  • 23 degree of freedom (thumb 1-2-2, finger 1-1-2-1)
  • 27 muscles
  • 17K cutaneous mechanoreceptors

• natural system architecture
  • mechanism : skeleton
  • transmission : tendons
  • actual : muscle
  • sensors : mechanoreceptors
  • control : motor neurons
• complex motor control
  • higher number of muscles and degree of freedom
  • neurological coupling between muscles
  • multi-articular muscles
  • redundancy
  • laws of mechanics : accelerate and then decelerate
  • coactivation : muscles groups active simultaneously, can happen in reciprocal activation or isolation
• brain function : know brain regions, don't know how they work together, no full model, hyptoheses based on experimental studies of particular motor task
  • three layers : activity (brain), commands (spinal cords), final movement (kinematic)
  • cortical organization : integrate visual, touch and perceptic information
• neurophysiological considerations : reaching, grasping
  • partiel cortex : visuomotor aspects
    • anterior intraparietal area AIP : response to viewing/grasping objects
  • ventral premotor cortex F5 : planning and execution, project finger motor neuron, fires according to kind of grasping, affordance
  • primary motor cortex F1 : dynamic aspects of movement, execute instructions sent by premotor cortex, generate real commands
  • area in/around intraparetal sulcus (VIP/MIP/LIP) : space representation used in F4
  • ventral premotor cortex F4 : planning, executing transport phase of grasph
  • in short
    • (VIP/MIP/LIP) - F4 - F1 pathway for reaching
    • AIP - F5 - F1 for grasping
  • human organization : similar to macaques
    • AIP : junction of anterior part of intraparietal sulcus and inferior postcentral sulcus
    • F5 : area 44
    • F4 : ventral part of area 6
  • CIPS : system input

• theories : possible theories

  • coding
    • motor cortex : high-level parameters (direction, magnitude of hand velocity)
    • motor cortex : low-level parameters (muscle force, mechanical power)
    • cortical neurons : force level generated by groups of muscles
    • direction of arm coded by neurons with varying fring rates as cosine function of angle between cell preferred direction and movement direction
  • planning
    • CNS plans multijoint arm movements using coordinates of joint angles and hand position, robust against inherent errors in motor system
    • cost function associated with production of movement trajectory
    • based local geometric features (two-third power law)
    • CNS selects particular movement primitives and rules so motor system resolve excess degree of freedom problem, defining number of solutions
    • major objective of motor coordination is achieving smoothest possible movement under circumstances
  • control
    • basic component of primate grasping : hand transport and preshaping
    • taken separately using two coordinated control mechanisms
    • taken as single limb using single control mechanisms
    • more plausible : coordinated smaller and simpler controllers
  • execution
    • hand velocity profiles : start and end characterized by bell shape
    • two-thirds power law : natural drawing consist of decreasing angular velocity with increasing curvature, $A=KC^{2/3}$ with $A$ angular velocity, $C$ movement curvature, $K$ constant

• models

  • internal model : neural mechanisms that mimics input-output relationships of limbs and external objects

    • forward internal : predict sensory consequences of given motor command, provide mechanism to overcome delay involved in sensing actual outcome, mental simulation, CNS use this for grip force modulation
    • inverse internal : starts from desired sensory state and outputs accordingly
    • grip force-load force coupling : observed when hand moved voluntarily while object held by finger opposing thumb, sensory delay in relaying change in load force to CNS 50-100ms

  • controls

    • FARS : neural ingredients of reach and grasp in terms of functional schemas, convert information to grasp affordances that are selected given task constraints, execution unfolds with a sequencing mechanism that monitor phases (maximum aperture reached, object contact) implemented by basal ganglia and presupplementary motor area F6 via inhibition and disinhibition

    • Hoff-Arbib : coordination of timing transport and preshape, time-to-completion signal, explain temporal relation of hand kinematics with finger aperture kinematics, modular decomposition of transport, preshape and enclose controllers, accounts for smooth correction in response to position or object size alteration, temporal invariance, schema that needs longer time slows the other down

    • Kawato : explain grip force-load force coupling, how CNS adjust grip force according to predicted using forward and inverse models, how CNS guarantees stability, 3 key computational elements arm controller, grip controller and forward model, good for single object

      

      • mosaic model : multiple objects, modular adaptive controller learn multiple dynamics with different properties, can implement Karato model

  • grasp learning : require two forces applied to object

    • pad opposition : between set of fingers, holding a peanut

    • palm opposition : between fingers and palm, holding a hammer

    • side opposition : between thumber volar surface and radial finger sides, holding a key

    • transform high degree of freedom into lower dimensional problem : which opposition for given object

    • framework : object properties perceived, object located in space, opposition determined, virtual fingers set up, hand aperture determined, grasp executed (preshape and enclose)

    • Iberall and Fagg : feedforward neural network to determine oppisitions and fingers

    • Oztop : developmentally oriented model, learn finger configuration for stable grasping

    • Uno : synthesize human like grasph using human motion capture data, produce right finger configuration, input visual information, output motor joint angles

• Manipolandum : upper limb, net stifness = sum stifness + stifness field

  • impedance control : resistance to imposed motion), might achieve stability, hard to show (null field vs divergent force field without require change in applied force), CNS voluntarily control magnitude, shape and orientation of endpoint stiffness in predictive way independent of force

• optimal control

• motor primitives

  • muscle synergies : reduce number muscle controlled, in principle orthogonal to each others, complex movements require as many syngergies as muscles
    • extraction : with activation coefficient
    • factorization : PCA, FA, ICA, NNMF
    • reconstruction
  • kinematic energy (recovery)

• study motor control : animal models, neuroimaging, TMS (transcranial magnetic stimulation), motion analysis, electrophysiological signal

  • muscle synergie : reduce complexity of motor control, extraction $D(t)=\sum_{i=1}^M c_i(t)w_i$ with $w$ synergies, $C$ coefficients, 5 synergies enough to explain upper limb variance
    • neuro-rehabilitation : modifies synergies

• redundancy : movement coordination master redundant degrees of freedom

  • manipulation motor control force production : individual fingers don't exert maximal force when working together (tendons originated from same muscle, a central neural drive (CND) has a max)

  • manipulation motor control posture : PCA for coarser level, finer level distributed among all joint angles

    

  • upper limb synergies : modification induced by disorders

Modifications in motor control of reaching and grasping due to neuropathologies

• prognosis : predicting likelihood of a person's survival/future
• stroke : interruption of blood supply to the brain, damage brain tissue
  • symptoms : sudden weakness unresponsiveness of the face, arm or leg, mostly on one side
    • spasticity : velocity dependent increase in tone, upper motor neuron dysfunction, abnormal regulation of strech reflex
    • apraxia : inability to perform learned task, motor planning areas MCA territory in dominant hemisphere
  • effect : post stroke
    • reach-to-grasp : variability, slower hand transport and aperture, inaccurate scaling of peak grasp, spatio-temporal coordindation decoupling, issue with extensor muscle activation, grip force less cordindates
    • upper limb : abnormal muscle coactivation as function of force direction and magnitude, reduced spatial and temporal efficiency, control using more feedback-guided with greater demand for error correction
    • rehabilitation : training towards patient specifc motor deficit, no learning deficit per se but slowness to develop required force to anticip control
    • quantitative motor control : differentiation between compensation and true recovery
  • recovery
    • (sub-)acute phase : first 6 months, try to recover healthy condition, use muscle synergies
    • chronic phase : after, compensatory movements, not able to learn back
• parkinson : progressive decline in motor function, proprioceptive deficits
  • symptomes : tremor, bradykinesia, rigidity and postural problems, loss of kinaesthetic sensitivity (conscious perception of limb and body motion) result in lack of precision and dexterity
    • dyskinesia : hypersensibility, might be improved by injecting L-Dopa
  • effect : dependent on external stimuli to initiate and shape motor output
    • reaching and grasping : distal body segments more affected, decreaserd movement speed, shorten amplitude, increased use of visual feedback, limited hand rotation, smaller peak aperture
      • reaching : always complete te movements, minimal rotatory, very little trunk movements (before limb, before reaching initiation, opposite direction of reaching)

Electrical stimulation for the restoration of grasping

• standard therapy : routine in hospital everyday

• electrical stimulation (ES) : short series of electrical current pulses between pairs of electodes, activate motor neurons

  • types : transcutaneous, percutaneous - though the skin, epimysial - implanted onto muscle surface, around nerve
  • pulses : monophasic or charge-balanced biphasic (symmertic or asymmetric - providing optimal control of contraction force whilst minimizing tissue damage)
  • fatigue : can reduce fatigue, not sure

• orthotics : modify structural or functional characteristices of muscles

• prostheses : grasp

  • HandMaster : array of 5 electrodes, control unit sends bi-phasic pulses, initiatve patterns

  • Bionic Glove : open loop with 4 superficial electrodes (3 stimulations with 1 reference), voluntary wrist flexion initiate stimulation

  • FreeHand : most advanced, 8 implanted epimysical electrodes and implanted stimulator, controlled by shoulder, high reliability

  • cuff electrodes : recruit several muscles, highly selective when possible, smoothness of recruitment questionable

• functional electical therapy FET : apply ES to clinical intervention for training, improve voluntary control of those muscles, acute stroke subjects benefit more

• implantable (intramuscular) systems for FET : electrode arrays, increased selectivty, simplify positioning problem, allow stimulation and EMG sensing, reduce preparation time, use for transradial amputee

• wearable systems for ES : customizable design patches, incresaed usability, integrate multimodal sensing and actuation, integration of different modules

• peripheral implants for ES

  • BIONic neurons : implantable, interface between electronic controllers and muscle, coil, surface stimulation tested first

• intraspinal stimulation : long trains of microstimulations evoke functional arm and hand movements in macaque, microwires

Neuroprosthetics

• history
  • rehabilitation : (re)integration of indivudual with disability into society, enhancing existing capabilities or providing alternative means
  • neurorehabilitation : recovery from nervous system injury, minimize or compensate for any functional alterations
    • 1920 : physiotherapy, motion therapy, gait therapy
    • use-dependent plasticity : cat with complete spina cord injury can stand and walk if practiced repetitevely, task specific, get detriment of other tasks
    • human patients : locomotor training improve recovery when lesion is incomplete, potential for plasticity decreases with age, assistance (passive, adaptive, active)
    • robot assisted neurorehabilitation : mistakes, no variability (key component to motor learning), learning helplessness, no active participation
    • exoskeleton : passive not that effective, lack variability and activity
    • soft exoskeleton
• gait rehabilitation platform
  • gravity : kinetic/potential energy transfert, cycle to cycle stability and reproducibility
    • gravity-assist : transparent mode, lift, forward-correction
    • transparent mode
• neuroprosthetics : developing neurl prostheses, series of devices than can substitute motor, sensory or cognitive modality that might have been damaged
  • neuromodulation : stimulation of nervous system to restore function
    • 18th century : Alessandro Volta stimulate auditory pathway, boom followed by boiling water
    • 19th century : microelectrode
    • 1963 : primitive brain radiocontrolled
  • neuroprosthesis
    • cochlear implants : works well as stimulating individual sensor, not the cortex
    • retinal neurprosthesis : patients "promised" to see again depressed by the results
    • pharmacological : monoamines modulate brain and spinal states, dopmaine replacement (parkinson)
    • neuromodulation : deep brain stimulation of basal ganglia, optogenetics
    • ​
  • brain machine interface : extract information from neural recordings to control devices, major issues stability of neuronal recording, of decoding algorithms and wireless embedded electronics
  • train spinal circuits : spinal cord injury, short term enabling activity, long term activity based plasticity, using modulation serotonin and glutamatergic excitation to revive neurons, 3D spinal cord reconstruction, nerves rewires
  • activity based training : enable short-term activity for further long-term plasticity
    • not very effective in human/rate : in case of severe contusion
    • need serotonin (modulation) and gluatamate (excitation) : provide resource to spinal cord to enable walking again
  • bright combinatorial future
    • next-generation : wireless neurostimulator, wireless spking activity, brain-spine interface, bi-directional neuroprosthetic systems
    • multifaceted : restoration (rehabilitation, modulation and robotics), replacement (BMI, artifical limbs, sensory feedback), marriage (brain to body and brain to brain interface)
• ecosystems : nervous system habitats
  • natural disaster : repair
    • stroke
    • sc1 : spinal cord
    • peripheral nerve injury
• ecoprosthetics : ecological strategies
  • recycle : rewire, plasticity triggered by training, engaging the patient
  • reduce : interaction with CNS, visibility, technologies (flexible electrodes, dura matter), personalized implants, avoid recalibration
  • reuse : do not change, reuse nerve in cut arm (autologous), available signal (EMG vs invasive), reinervation (e.g. muscle not used except body builder)
  • rescue : gene therapy, affect spinal cord by targetting motor cortex
• restauration : mental flow
• rehabilitation : reliable long-term
• stem cells : die in healthy, need to be on injuries border, need to come from same region (spinal cord for spinal cord)
• alzeihmer : might be protein staturation before neuron loss
• feedback : hard to be used daily, pump up amplitude to enforce stronger feeling (harder touch)

Hand prostheses

• motivation : quality of life, reaching, grasping, manipulation, sensing, gesture, communication

• big challenges : functional and natural controlled prosthetics hand (dexterity, reliability, cognitive effort)

• prosthetics hands

  • passive : cosmetic
  • active : body powered, myolectric

• state of the art : poor functionality (limit set of grasp, high cognitive efort), poor cosmesis (static and dynamics properties), poor controllability (no sensory feedback), 30-50% users abandon their prosthesis

• mechatronic research

  • intelligent mechanisms (main issues) : non-back-drivability, adaptive mechanism, allow different prehensible patterns (manipulation)
  • sensors & their integration : low-level automatic control, feedback delivery
  • intelligent control : execute automatic loops, deal with interface

• design issues

  • adaptability : hard design, actuate and control enough degree of freedom, solutions include simplification or adaptable mechanisms
  • non back drivability : hold grasp once power switched off, power saving

• SMART hand prototype : reliability, grasph capabilities, cosmesis

• OPENHAND prototype : multi-layers structures (edidermis as compilant materials, nail and bones as stiff materials), non-linear and time-dependent characteristic

• user-prosthesis interface

  • non-invasive
    • bio-signal interfaces : EMG, cortical BCI (EEG)
    • sensory subsitution feedback systems : electrotactile, vibrotactile, pressure, auditive
    • others : foot, tongue, speech, EOG (eye)
  • low invasive : targeted muscle reinnervation (TMR), implantable myoelectric sensors (IMES)
  • invasive : direct neural control (ECoG, LFP), CNS, PNS

• sensing the brain : TIME vs SPACE invasiveness tradeoff

  • surface EMG : not-homologous voluntary movements coded as prosthesis movements, require long training, hard to control more than two degrees of freedom, cheap

  • IMES : sense myoeletrical signal at source, BION

  • electrodes

    • TIME : transveral intrafascicular multichannel electrodes, selective

      

  • TMR : sensory feedback possible but difficult to be daily usable

  • sensory feedback : intracortical performance limited, reuse existing neural structures when possible, closed-loop control allow user to achieve performance close to natural ones

• texture recognition

• embodiment : thoughts, emotions, behaviour based on sensori experiments and positions

• proprioception : sense of limb position and movements, use sensory re-mapping in real-time in conjuntion with tactile feedback

• quest for bionic hand

• i4LIFE