VestibuloOcularReflex

links:

• Motor Coding in Floccular Climbing fibers
• The site of a motor memory shifts with consolidation
• Learning in a simple motor system (February 2006) Broussard DM, Kassardjian CD. Learning in a simple motor system. Learn Mem. 2004 Mar-Apr;11(2):127-36.
  • due to its pivotal nature, motor learning may have been one of the first forms of learning to be implemented by biology. "universal in freely-moving animals"
  • these authors define it as procedural - does not require concious attention (but it can be influenced by it)
  • eye movements procedurally simpler than arm movements, which I've spent some time looking at.
    • lately, I've been having to do a lot of this, after my glasses lost one earpeice and have been moving about since :) slight changes in the angles of the lenses are very noticeable.
  • I thought that the theory for this system would be complete by now, and ready for the application to more complicated motor movements, but this is not so. The thoeries are still a bit controversial, and may require a molecular understanding.
  • the general principle behind the error hypothesis is pretty simple: whatever caused the error was bad, and therefore the weights of those inputs that occur subsequently should be decreased. However, that is only the tiniest slice of the puzzle.

Description of the system:

• see figure 1 of learning and memory. labels would be very useful. nb the lateral rectus muscle is innervated by the abducens (cranial nerve VI)
• need diagram of the direction that the eyes turn, including the lateral/medial rectus muscles.
  • Note that the eyes are stabilized in the two other directions - pitch and roll. here we study yaw, but people have demonstrated the same effects in the other directions.
  • { cat and human have gain < 1, monkey just about 1 (perfect) - humans require extra input, via OKR}
• basic circuit known since 1967 (Eccles).
• Maekawa and Simpson: the cerebelar purkinje cell recieves climbing fiber input from the inferior olivary nucleus that encodes visual information.
• information about ongoing movements arrives at the lateral vestibulocerebellum via mossy fibers from the dorsolateral pontine nuclei.
• some mossy fibers also carry visual information.
• purkinje cells project directly back to the vestibular nuclei.
• (initial) adaptation is fast, 15 minutes in humans and monkeys
  • * * for lateral-eyed animals, lenses can be replaced by a rotating drum with a high-contrast pattern printed on the inside. Drum is moved based on the animals head motion - whether voluntary or involuntary. Humans and monkeys, with stereo vision, require lenses.
• this structure is supported by the timescale/waveshape of the output adaptation, which features a fixed, fast compensation followed by a slightly slower, adaptive response (figure from sejnowski).
• required elements:

1. cerebellar cortex
   1. inactivation of PKC in purkinje cells of the floculus abolished optical VOR adaptation in mice.
2. vision plus another signal, which can be vestibular or oculomotor (eg the output of the system itself)
3. a reason to change, injury, prism glasses, magnifying spectacles, etc.

• ito's hypothesis:
  • climbing fibers froms the inferior olive provide a visual teacher signal that adjusts the sensitivity of the purkinje cell to particular inputs from parallel fibers.
  • algorithm:
    • all movement viewed from the top. cw = positive. image movement on the retina: to the right = positive.
    • CW rotation of the head it is important to pay attention to sign! increases the firing rate of Vestibular neurons (HSSC) on the right side.
    • exitatory synapes onto the inhibitory interneurons
    • -> decreased activity in the ocular motoneurons
    • -> lateral rectus is relaxed
    • -> eye rotates counterclockwise.
  • ok, now say the eye rotates too little! (negative slip, want to increse the gain)
    • -> activity in inferior olivary nucleus is topologically organized, increases in activity it is equally important to pay attention to signal label. It seems to be that label here is determined (partially) by genetics, though not completely because humans are able to conciously change their VOR.
    • -> concurrent activity of the climbing fibers and mossy fibers from the vestibular nucleus
    • -> causes LTD changes in the excitatory PF-Purkinje synapse
    • -> resulting in less PK activity,
    • -> inhibiting the interneuron less - making the interneuron more exited
    • -> causing the lateral rectus muscle to have greater sensitivity & fix the problem.
  • generally, the gain will increase if (and decrease, vice-vercia)
    • the head movement and retinal slip are in opposite directions
    • the eye movement and retinal slip are in the same direction.
      • smooth pursuit alone, against a dark background (no vision) can change the gain of the VOR.
      • in other words, parafoveal retinal slip with eye movement in the same direction can increas VOR gain.
• timescale of algorithm is consistent with gradient-descent learning (expodential).
• Miles et. al 1980 - competing hypothesis to ito - purkinje cells decrease their sensitivity to rotation (vestibular) when VOR gain decreases. This doesn't make all that much sense! when the PC decreases its sensitivity, the inhibitory interneuron increases sensitivity, and so there is more response in the muscle.
• VOR interneurons modify their responses dramatically during lerning - they can reverse the direction of their sensitivity. These neurons project to the ipsilateral abducens nucleus ; the exitatory, contralateral neurons probably remain unmodified.
• models predict that the system is unstable if there is only a single site of plasticity.
• the memory is apparently consolodated elsewhere
  • shutdown of the cerebelum via tetanization of the climbing fibers is insufficient to abolish a learned VOR response .
  • did have an effect when lidocane was injected into the floculus of goldfish (this is a really ancient system!!)
  • in the long term, the cerebellum contributes less to the storage of motor memory than it does in the short term. (long-term changes in the vestibular nuclei, in accord with Sejnowski PNP cells)
• OKR to higher frequencies is rather low - < 10Hz, as confirmed by Winkelman & Frens - and is insufficient when the animal is moving quickly. The stabilization relies on the vestibular nuclei, which are fast (14ms - 19ms).
  • -> despite this learning requires vision or at least movement - the error signal is conditional on the slow data available from vision
  • !! counterpoint:
    • error signals other than retinal slip can drive adaptation (can track in strobe light, using after-images).
    • saccades also drive adaptation.
• Vestibular Nuclei page -- great resource (for both pictures and text).
  • The vestibular nuclei receive their primary input from the vestibular portion of C.N. VIII (8, vestibular-auditory)
  • Vestibular Nuclei receives information from the receptors of the vestibular labyrinth, i.e. hair cells located in the semicircular canals and the saccule and the utricle (otolith organs).
• trying to understand the cerebellum well enough to build one
• Neural learning rules for the vestibulo-ocular reflex
• vestibular adaptation in the lab --includes some interesting controls where the subject imagined saccades and movements!
• Cerebellum-dependent learning: the role of multiple platicity mechanisms Cerebellum-Dependent Learning: The Role of Multiple Plasticity Mechanisms Edward S. Boyden ES , ­Akira Katoh A, and ­Jennifer L. Raymond JL. Annual Review of Neuroscience Vol. 27: 581-609 July 2004.
  • feat. a decent diagram and plenty of text on the nature of generalization and sparse coding (I haven't read this yet).
• Eyes on Target Angelaki DE. Eyes on target: what neurons must do for the vestibuloocular reflex during linear motion. J Neurophysiol. 2004 Jul;92(1):20-35.
• Ito - The Molecular organization of cerebellar long-term depression Ito M. The molecular organization of cerebellar long-term depression. Nat Rev Neurosci. 2002 Nov;3(11):896-902. -- another nice pic here, includes the red nucleus in the diagram, but who knows what that does. certanly looks like efferent motor command...
• Eggers et al 2003 Non-visual error source of VOR Eggers SD, De Pennington N, Walker MF, Shelhamer M, Zee DS. Short-term adaptation of the VOR: non-retinal-slip error signals and saccade substitution.
  • position and velocity gain can be trained independently.
  • position gain is sensitive / can be trained by corrective saccades. (short-term 30 minutes, 15 deg head jolt - eg. in the cerebellum)
• Internal models in the cerebellum Wolpert, D.M., Miall, R.C., & Kawato, M. (1998). Internal models in the cerebellum. Trends Cogn. Sci. - interesting, plenty of control-flow diagrams :) <pre> Acquiring an inverse dynamics model through motor learning is generally a difficult task because the error in the model’s output, the motor command error, which could provide a training signal is not directly available to the CNS. If the motor command error was known, there would be no need to learn the inverse dynamics as the correct control signal would already be known. Instead movement errors are initially represented in sensory coordinates, and these sensory errors need to be converted into motor errors before they can be used to train an inverse model. </pre>
  • concerning their model: Because the inverse model possesses input–output transfer characteristics that are the inverse of those of the controlled object, the cascade of the two systems gives an approximate identity function.
• Input Minimization - a model of cerebellar learning without climbing fiber error signals Neuro Report ?? . Poses an alternative to the widely-accepted error (based) Marr and Albus theory, in which the complex spikes are used for learning. unlikely to be true, the complex spikes are too different to not matter, and LTD is too robust and essential in leaning.
• One cannot build theories of cerebellar function on shaky foundations: Induction properties of long-term depression have to be taken into account
  • temporal credit assignment problem : It has been shown both in decerebrated animals and in slice that climbing fiber stimulation must precede parallel fiber stimulation to obtain LTD induction (Ekerot & Kano 1989; Schreurs & Alkon 1993; Karachot et al. 1994)
    • This is impossible to reconcile with the proposed role of the climbing fiber as an error signal
    • -> well, not really. so long as you assume the latency of CS is less than 1/bandwidth of signal you are trying to track, you will be able to propagate a gradient-descent error.
• On climbing fiber signals and their consequences Simpson JI; Wylie DR.; De Zeeuw CI. More on climbing fiber signals and their consequence(s) .Behavioral & brain sciences.1996;19:496
  • climbing fibers
    • all climbing fibers originate from the contralateral inferior olive
    • each purkinje cell synapses with one climbing fiber
    • climbing fibers fire infrequently (1-2Hz), but can synchronize a number of purkinje cells through their strong synapses. some people think this synchronization is important. it probably is.
  • Purkinje cells within a given zone project to only one cerebellar nucleus, which receives collaterals from the inferior olivary axons that terminate as climbing fibers in that zone
  • Furthermore, the cerebellar nuclei contain GABAergic neurons that project contralaterally to those parts of the inferior olive (De Zeeuw et al., 1989; Nelson & Mugnaini, 1989) that provide the collaterals to that particular cerebellar nucleus (Dom et al., 1973; Graybiel et al., 1973).
  • this is called a module, and the climbing fibers are essential in parceling the cerebellum into modules.
  • olivary neurons
    • inferior olibe is perhaps the largest structure in the brainstem.. and still quite mysterious.
    • olivary neurons discharge either a single spike or a burst of spikes (2-5 spikes with an interspike interval of 2-3 milliseconds) about once or twice per second (Armstrong, 1974; Armstrong & Rawson, 1979; Crill, 1970)
    • olivary cells fire rhythmically due to the nature of their cell membrane conductances.
    • olivary cells tend to fire synchronously due to the fact that they are electrotonically coupled by dendrodendritic gap junctions. this results in <2ms synchrony in (measured) 40% synchrony in the the climbing fibers of a single cerebellar zone. This is also true when the recorded purkinje cells are 1mm apart (! try that in the cerebrum )
    • Takeda T, Maekawa K. Bilateral visual inputs to the dorsal cap of inferior olive: differential localization and inhibitory interactions. Exp Brain Res. 1980;39(4):461-71.
      • two pathways known to mediate optic signals to the rabbit flocculus, from ipsilateral (projects into rostral part) and contralateral (caudal) eyes. CF subsequently cross over again?
      • there is inhibition between the 2 pathways.
    • Haddad GM, Demer JL, Robinson DA. The effect of lesions of the dorsal cap of the inferior olive on the vestibulo-ocular and optokinetic systems of the cat. Brain Res. 1980 Mar 10;185(2):265-75.
      • removal of the dorsal cap of the inferior olive blocks prism reversal
      • but it does not totally abolish VOR adaptation.
  • in contrast - as we know - Purkinje cells fire simple spikes very frequently, ~100Hz, due to exitatory parallel fiber input.
  • Purkinje cells pause in their output of simple spikes 10-100ms after a complex spike; others gradually fade back to normal spiking; others transiently increase their spiking. They sem to fall into three distinct classes. importance?? possibly an artifact. small modulation is insufficient to explain changed motor output.
  • consequences of complex spikes are somewhat better known than the actual information content of the events. (which is what Winkelman & Frens looked at).
  • the flocculus
    • five zones (1, 2, 3, 4 and C2) whose borders can be delineated in the floccular white matter by using acetylcholinesterase (AChE) staining
    • climbing fiber projections to zones 1 and 3 are derived from the rostral dorsal cap and ventrolateral outgrowth;
    • climbing fiber projections to zones 2 and 4 are derived from the caudal dorsal cap
    • climbing fiber projection to zone C2 is derived from the rostral pole of the medial accessory olive
    • In zones 1-4 the complex spikes are optimally modulated by rotational optokinetic stimulation about either the vertical axis or about a horizontal axis approximately perpendicular to the plane of the ipsilateral anterior semicircular canal. These optimal axes have a geometry similar to that of the best-response axes of the semicircular canals and to the axes about which the three pairs of extraocular muscles rotate the eye.
    • complex spikes divided into two classes based on ocular dominance.
      • those dominated by the contralateral eye respond best to rotation about an axis at about 45 contralateral azimuth;
      • those dominated by the ipsilateral eye respond best to rotation about an axis at about 135 ipsilateral
      • vertical axis (VA) neurons are located in zones 2 and 4, while the contra-45 and ipsi-135 neurons are located in zones 1 and 3.
    • the climbing fibers of zone C2 do not respond to optokinetic stimulation.
    • Collateralization of climbing and mossy fibers projecting to the nodulus and flocculus of the rat cerebellum Ruigrok TJ. Collateralization of climbing and mossy fibers projecting to the nodulus and flocculus of the rat cerebellum. J Comp Neurol. 2003 Nov 10;466(2):278-98.
  • there are visually selective climbing fibers in the flocculonodular lobe - these originate in the dorsal cap of Kooy and the ventolateral outgrowth, and signal the direction and speed of movement of large parts of the visual world across the retina.
  • there are vestibularaly active neurons in the beta nucleus of the inferior olive, which sometimes respond tonically to head tilt.
  • Gellman et al. (1985) found in the awake cat that complex spike responses to a passively applied stimulus generally failed to occur when a similar stimulus was produced by a voluntary movement, unless the receptive field of the complex spike was "unexpectedly" brought into contact with an object during active movement
  • the overall nature of LTD is, at the core, simple: whatever caused the error was bad, and therefore the weights of those inputs that occur subsequently should be decreased. The problem, again, is causality, or temporal credit assignment, and may be effectively solved with the smith-model, or by simply repeating neural behavior to properly assign a punishment. <pre>That is, reduction of the simple spike activity of a Purkinje cell that receives from a given climbing fiber should lead to a sensory input that opposes the occurrence of that climbing fiber's discharge </pre>
  • LTD has a strong dependence on frequency - low levels of CF activity is insufficient to initiate it. Ito 2002 Ito M. Mechanisms of motor learning in the cerebellum. Brain Res. 2000 Dec 15;886(1-2):237-245.
  • some climbing fibers are very, very sensitive to tactile stimulation of the foot-paw.
  • how does our world remain stable with incessant climbing fiber activity? memory (personality, etc) seems so permanent, but its substrate may be volatile. Trace this back to Neuronal correlates of movement dynamics in the dorsal and ventral premotor area in the monkey which shows that after training, (a small fraction of) neurons Do Not resume their original coding. NB Received: 2 October 2003 Accepted: 18 May 2005 Published online: 22 September 2005
  • random: Lin has noted that human reaction-time movements have been reported to be paced by a normal 10 Hz physiological tremor
• The Oculomotor System: Anatomy & Physiology - good diagrams.
• Learning and memory in the vestibulo-ocular reflex S du Lac, JL Raymond, TJ Sejnowski, SG Lisberger. Learning and Memory in the Vestibulo-Ocular Reflex - An. Rev. of Neuroscience, Vol. 18: 409-441 1995.
• Historical review of the significance of the cerebellum and the role of Purkinje cells in motor learning Ito M. Historical review of the significance of the cerebellum and the role of Purkinje cells in motor learning. Ann N Y Acad Sci. 2002 Dec;978:273-88.

$$A\sin (t+v)=\sin \left(t\right)\cos \left(v\right)+\cos \left(t\right)\sin \left(v\right)$$ $$A_1\sin (t+{\theta }_1)+A_2\sin (u+{\theta }_2)=\sin \left(t\right)\left[A_1\cos \right({\theta }_1)+A_2\cos ({\theta }_2\left)\right]+\cos \left(t\right)\left[A_1\sin \right({\theta }_1)+A_2\sin ({\theta }_2\left)\right]$$

$$T(\mathrm{cs}\mid x)=\sum _{i=1}^{N_{\mathrm{spikes}}}p(x_i\mid {\mathrm{cs}}_i){\log }_2\left(\frac{p(x_i\mid \mathrm{cs})}{p\left(x\right)}\right)$$