The Functional Anatomy of the Semicircular Canals in Man
The semicircular canals are generally described as a set ofmutually perpendicular membranous rings that meet in a dilated expansion calledthe utricle. The anterior andposterior canals are approximately vertical with each canal at an approximately45” angle to the midsagittal plane and the lateral canal approximatelyhorizontal, which is why it is also called the horizontal canal.
The vestibular part of the membranous labyrinth iscomposed of the utricle and saccule and three roughly mutually perpendicularsemicircular canals. Thesemicircular canals arise from and end in the utricle.
The semicircular canals lie in a similarly shaped bonychamber called the bony labyrinth, which also has a cochlea and threesemicircular canals. The canals,along with the utricle and saccule and the scala media of the cochlea and theendolymphatic duct form a complex membranous structure called the membranouslabyrinth. The scala media isattached to the walls of the bony cochlea and the membranous semicircularcanals lie within the bony semicircular canals. The utricle and saccule lie in the vestibule of the bonylabyrinth.
The membranous labyrinth is filled with an ionic solutioncalled endolymph and surrounded by another ionic fluid called perilymph.
The vestibular parts of the membranous labyrinth, thesaccule and utricle and the three semicircular canals are suspended in the bonylabyrinth by thin connective tissue trabeculae, so that the membranouslabyrinth is floating in the bony labyrinth, but retrained from large movementswithin the bony chamber. Thearrangement is like a water-filled balloon floating in a substantially largerpool of water, held in place by many thin elastic bands.
The cross-sectional radius of the membranous semicircularcanal is about one-third that of the bony labyrinth and it is modestlyflattened so that the canal has an elliptical cross-section.
At one end of each canal, just before it meets the utricle,the membranous semicircular canal dilates into a bulbous ending called theampulla of the canal. The ampullahas about three times the radial cross-section of the remainder of thecanal. The ampulla has a transverseridge, called the crista ampullae, that contains sensory hair cells oriented soas to respond head movements in the plane of the canal.
Nerves and blood vessels that pass into the base of thecrista supply the sensory epithelium of the crista ampullae.
Rotation of the head in the plane of a canal causes a slightlag in the flow of the endolymphatic fluid within the canals because of itsinertia. The forces involved aremuch the same as you experience when you are in a car that accelerates ordecelerated or turns a corner. Your body would continue in the same direction at the same speed exceptthat it is compelled to change its velocity by your seat pressing on yourbody. In a semicircular canal, itis the walls of the membranous canal and ultimately the bony canal that compelmovement. The fluid can also movearound the canal if there is not a wall to force it in the new direction or atthe new speed. Lagging fluid moving within the membranous canal ultimatelydisplaces cupula of the ampulla, causing it to billow like a sail in abreeze. As the cupula billows inthe displaced endolymphatic fluid, the cilia of the embedded hair cells aredisplaced.
Displacement of the many stereocilia of a hair cell towardsthe single kinocilium at the margin of the cluster of stereocilia will causechannels to open, leading to a depolarization of the hair cell.
Each canal has a dilation at one end where it meets theutricle. That dilation, called theampulla has a transverse ridge, called the crista ampullae, that contains apopulation of embedded hair cells, all oriented so that displacement of thehairs in one direction elicits depolarization of the hair cells and movement inthe opposite direction causes hyperpolarization. The crista ampullae is surmounted by a gelatinous membrane,called the cupula, which seals off the ampulla. Movements of the fluid in asemicircular canal causes the cupula to billow, like a sail in a breeze, andthe shearing force exerted upon the embedded stereocilia and kinocila of thehair cells causes more or less neurotransmitter to be released on to theterminal processes of the sensory neurons of the vestibular nerve.
Rotation in one direction will cause the discharge rates toincrease and rotation in the opposite direction will cause them todecrease. Rotation in planes thatare oblique to the canal plane will have less and less effect, until rotationsin a plane perpendicular to the canal plane will not modify activity in thecanalÕs hair cells. Since thereare three canals that are approximately perpendicular to each other, one wouldexpect that rotations of the head in any direction will stimulate at least onecanal and most often multiple canals.
The mirror symmetrical arrangement of the canals on the twosides of the head means that any rotation will stimulate canals on both sidesof the head. The two sets ofcanals are set so that the two horizontal canals respond to rotations inopposite directions. The anteriorcanal on one side is approximately aligned with the posterior canal on theopposite side so rotations in certain oblique planes will affect a canal on oneside and the other canal on the opposite side. For instance, rotating the head so that the nose moves downand to the side will increase activity in the opposite anterior canal anddecrease activity in the posterior canal on the same side.
We will consider the sensitivities of the semicircularcanals in terms of vectors that are orthogonal to the planes of thecanals. This approach is generallyconsidered more difficult that using the planes of the canal themselves, but,with a little practice, one soon realizes that normal vectors are easier tounderstand, because they lack the ambiguity intrinsic to the planar approachand the magnitude and direction of a vector is easier to picture than theorientation of a plane. Inpractice, it will be convenient to move back and forth between the twodescriptions of the canals. Afterall, the planes are what we see when we look at the canals and rotation in theplane of the canal is intuitively clear, if difficult to quantify.
If we consider the sensitivity of the anterior canal, acanal at a roughly 45” angle to a sagittal plane, movements that cause theampulla to move anteriorly will excite hair cells in the crista ampullae of the canal.
Briefly consider the movements of the nose and ear duringoptimal movements for exciting the anterior canal, where the axis passesthrough or near the semicircular canal. The nose moves down and towards the opposite side from the canal, while theear moves down and towards the same side. This disparity is due to the geometry of the head and it is only anapparent difference. If the axisof rotation is set higher, then both the nose and the ear move to the oppositeside and, if it is set lower, they both move to the same side.
Each canal has two, oppositely directed, normal vectorswhich are perpendicular to the plane of the canal and which can be an optimalaxis of rotation. Semicircularcanals are stimulated by angular momentum and angular momentum is expressed bya vector perpendicular to the plane of rotation, that is, the plane defined bythe movement vector and a vector that extends perpendicularly from the axis ofrotation to the moving mass. Forthe most effective stimulation of a semicircular canal the center of rotationis on an axis of rotation perpendicular to the plane of the canal where itpenetrates the plane of the canal. Therefore, either normal vector to a canal is potentially the optimalaxis of rotation for the canal. However, rotation in one direction will depolarize the hair cells in thecanalÕs ampulla and rotation in the opposite direction will hyperpolarize them.
If one has a plane that includes the origin, it may bedescribed by an expression like the following.
The are constantsand the x, y, and z are variables. Normally, the plane is written in the form of the second expression, butthe basis vectors in the first expression are implied.
If we express the plane in this format, then it is apparentthat a plane is the set of vectors for which the sum of their coefficients iszero. The orientation of the planeis determined by the relative values of the constants.
We can readily calculate two vectors that lie in the planeby setting the coefficient of k equalto zero and setting the coefficient of j equal to zero.
Then the normal vector to the plane is the vector of theratio of these two vectors in the plane. We can take the ratio of the second to the first or the first to thesecond.
If we compute the ratio of the second to the first, then weget a vector that extends in one direction relative to the plane and if wecompute the other ratio, then the vector points in the opposite direction.
Because of this relationship between the expressions for aplane and its unit normal vectors, we can take descriptions of the canal planesand immediately write down the expressions for the axes of rotation for thecanals.
Blanks et al.(1975) measured the planes of the semicircular canals in ten skulls of amixture of races and they expressed the planes of the canals as the regressionsof all of the canals of a given type. We can directly convert their expression for the regression plane intothe axes of rotation for each canal type. For the canals on the right side the values were as follows.
Horizontal canal | CHR | 0.365 i + 0.158 j - 0.905 k |
Anterior canal | CAR | 0.652 i + 0.753 j + 0.17 k |
Posterior canal | CPR | 0.757 i – 0.561 j + 0.320 k |
To get the values for the canals on the left side, we needonly reverse the values for the j terms.
Horizontal canal | CHR | 0.365 i - 0.158 j - 0.905 k |
Anterior canal | CAR | 0.652 i - 0.753 j + 0.17 k |
Posterior canal | CPR | 0.757 i + 0.561 j + 0.320 k |
However, a little thought and experimentation will convinceyou that the values as written are for movements relative to a left-handedcoordinate system. To render theminto a right handed coordinate system one must reverse the values for all thecomponents of the left-handed expressions. So the final values for the left-handed canals are as follows.
Horizontal canal | CHR | - 0.365 i + 0.158 j + 0.905 k |
Anterior canal | CAR | - 0.652 i + 0.753 j - 0.17 k |
Posterior canal | CPR | - 0.757 i - 0.561 j - 0.320 k |
It is much easier to think about the canal axes when one hasa graphical representation. Tothat end, this section considers a number of images that represent the canalaxes as vectors. In the firstfigure the canal axes for the semicircular canals on the right side of the headare represented as viewed from two viewpoints. The left panel is an oblique view from the left and abovethe horizontal plane and the right panel is the same axes viewed from directlyin front of the head and slightly above the horizontal plane.
One can see that the horizontal canal axis is verticallydirected with a modest anterior and lateral tilt. Its total tilt from vertical is 23.7”; it tilts 10”laterally and 22” anteriorly. Theanterior canal has an axis that crosses the midline at a roughly 45” angle to asagittal plane, actually 49.11” to the contralateral side a sagittalplane. It is nearly confined to ahorizontal plane, being 1” below the horizontal plane.
All these measurements are relative to the Reid stereotaxiccoordinates in which the horizontal plane passes through the centers of theexternal auditory canals and the inferior margins of the orbits.
The right semicircular canal axes viewed from theleft and anterior (left panel) and from in front and slightly above thehorizontal plane (right panel). The basisvectors are aligned with the anterior axis (i), the left lateral axis (j), and the vertical axis (k) of the Reid stereotaxic system (Blanks et al,Õ75). The canals axes are labeledwith H for the horizontalcanal, A for the anteriorcanal, and P for theposterior canal. These conventionsapply to all of the following figures unless stated otherwise.
The next task is to represent the relationships between thecanal axes on the two sides of the head. In the next figure the right canal axes have been reflected in asagittal plane, because the canals on the two sides of the head areapproximately mirror reflections of each other. That representation is quite intuitive in that one set ofcanal axes is simply the reflection of the other set. Unfortunately, reflection also flips the coordinatesystem. A bit of thought andexperimentation will reveal that the axes on the left side are appropriate onlyif one uses oneÕs left hand to express the rotation. They are in a left-handed coordinate system, while the axeson the right side are in a right-handed coordinate system.
We will consider the reflection of the right canal axes in abit, but there are some points that can be drawn from this representation thatare not as easily appreciate in a consistent single-handed representation.
The right and left canal axes with the rightcanal axes reflected across the midline. This is anintuitive representation, but a little experimentation will reveal that theleft side axes work only if one uses oneÕs left hand to indicate the movement,that is they are in a left-handed coordinate system while the ones on the rightare in a right-handed coordinate system. The right canal axes have an Ras a subscript and the left canal axes have an L.
The next figure is similar to the last figure except that thecanal planes have been represented by circular discs that are perpendicular tothe axes of rotation. The centersof the canal axes have also been moved a bit further apart so that the anteriorcanal axes do not overlap and the basis vectors are clearly visible.
The reflected canal axes are drawn with theplanes of the canals indicated by circular discs and the canal centers moved further apart, so thatthe anterior canal axes do not overlap. The same reservations apply to this representation as for the previousfigure. One can see that the canalaxes do not leave through the intersections of the discs, which indicates thatthe canals are not actually orthogonal, although they are approximately so.
As stated above, the canal axes in the previous two figuresare in two different coordinate systems, which aids in understanding therelationships between the canals on the two sides of the head, but they willrapidly get one in trouble if one tries to compute with them.
The canal axes expressed in right-handedcoordinate system.
As with the reflection that changed the handedness of thecanal axes, one can see that the three canals on one side are approximatelymutually perpendicular, but deviate from perpendicular by a readily detectedamount. The vertical canals onlyapproximately align with the matching canal on the opposite side (rightposterior with left anterior and right anterior with left posterior).
The next three figures show the same canal axes from anumber of special viewpoints. Thefirst is from a point directly anterior to the canals.
The symmetry of the two sets of canal axes is a more subtlesymmetry than one normally perceives, but they are truly symmetrical becausethey are transformed into each other with only a reversal of signs for thedirectional components. We alsoknow that the representations are reflections of anatomical structures that aresymmetrical across the midsagittal plane.
The same canal axes viewed from directlyanterior.
The same canal axes viewed from directlylaterally.
The same canal axes viewed from directly above.
Given the plane and normal vectors of the canals, one canreadily compute the relative orientations of the canals.
Canals | Ratio | Angle | Intersection | Blanks et al. Ō75 |
Horizontal to Anterior | | 112.216” | | |
Horizontal to Posterior | | 95.948” | | |
Anterior to Posterior | | 86.2003” | | |
Right Posterior to Left Anterior | | 23.2833” | | |
Right Anterior to Left Posterior | | 23.2833” | | |
Right Horizontal to Left Horizontal | | 18.3942” | | |
Most movements will not be about an optimal canal axis forany canal and no rotation can be optimal for all the canals, because they havedifferent canal axes. That isfundamentally why there are multiple canals, so that any head rotation can besensed by at least one canal. Wenow consider how the different canals sense a given rotation.
The magnitude of response in a canal is proportional to theprojection of the axis of the rotation on the canal axis.
The canal axis is a unit vector, so, the response is avector in the direction of the canal axis that varies about the resting rate toan extent proportional to the magnitude of the movement, the proportionalityfactor, and the projection of the axis of rotation upon the canal axis.
In this expression the summation is over canals, althoughone could view it as a summation over canal afferents.
If the nervous system responds to the movement, then itneeds to know the current direction of the movement and its rate of change withrespect to time.
All the canal axes are drawn from a common centerand viewed from the left front, above the horizontal plane.
Cupula as a diaphragm between a large, low resistance,chamber and a small, high resistance, chamber
The conformational change of the flattened semicircular ductas a short time constant capacitance/spring.
Short duration signals affect primarily the capacitance
Long duration signals affect the resistive elements
The dynamic vestibular response behaves like a second degreedifferential equation with a very short time constant (~1/100 sec) and a longtime constant (20-30 sec).
The stimulus is almost certainly rotational head movements,which cause angular acceleration.
The short-term response mimics the angular velocity, whichis taken to indicate a mechanical integration.
The long-term response is characterized by an adaptivereturn to baseline in the presence of a sustained constant angular velocity.
The model is generally characterized by a driving functionthat is a torque, therefore proportional to the second derivative of an angulardisplacement. The semicircularduct responds with an inertial drag from the fluid within the duct, resistanceto flow of the fluid because of viscosity (proportional to first derivative ofangular displacement), and distortion of the cupular membrane as it resists therelative displacement of the fluid (proportional to angular displacement).
The semicircular duct is quite narrow, which results in asubstantial resistance to flow along its axis, and it is flattened, which meansthat has a cross-sectional area that is less than maximal for itscircumference. If it is necessaryfor fluid to flow through the duct, there will be a resistance that will bepartially offset by outwardly directed pressure on the duct walls, which willcause the duct to become more circular in cross-section.
A pressure difference across the cupula will cause thecupula to billow like a sail in a breeze. Since the cupula is elastic, it will provide a restorative force thatresists fluid flow and returns the fluid to status quo
As the cupula billows it produces a transient pressure changesin the ampulla and the proximal duct. These are partially absorbed by fluid flow through the duct andpartially by changing the ductÕs cross-sectional shape to accommodate more orless fluid. The resistance tofluid flow through the duct is proportional to the velocity of flow.
Because of the much greater resistance of the duct to fluidflow, one can open the circuit and to the first approximation model thecupular-duct dynamics as two chambers with an elastic diaphragm between.
The forces on the cupula will be balanced so we can write anumber of expressions
Let represent thedisplaced volume. On the utricularside of the cupula there is a force that causes aflow through a resistance .
The elastic force, , from distortion of the cupula is linearly related to thedisplaced volume
The duct force is composed of the conformational distortion,, and the resistance to flow, . Theconformational distortion is proportional to the displaced volume for smallvolumes.
The flow resistance is proportional to the rate of change ofthe displaced volume or the flow.
If the duct is fed back to the large reservoir then therewill be a different force applied to its distal end, .
The two forces that impinge upon the ends of thesemicircular ducts, , are dependent upon the geometry of the system and themanner in which it moves.
The torsion pendulum model assumes that rotations occuraround the center of the ring, but they almost never do so.
An axis drawn from the foramen magnum to the semicircularcanals would be roughly orthogonal to the posterior canal, roughly in the planeof the anterior canal and roughly in the plane of the horizontal canal.
The positioning of the cupula at the utriculo-ampullaryjunction may important to the changes that occur in the cupula.
The sensory apparatus for a semicircular canal is located inthe ampulla, so, it is movements of the ampulla that are most relevant to thestimulation of a canal. For thatreason it makes sense to place the framework for describing the canal at theampulla. The framework will be astandard frame of reference. Thelocation of the ampulla, , is the first frame vector. It will be defined relative to a central origin,
The framed vector for a canal is anchored at theampulla and its orientation reflects the movements that maximally stimulate theampulla.
Movements in the direction of or
Canal afferents have a rest discharge rate that may increaseor decrease, depending on the direction of head movement.
The magnitude of the movement has been separated from thedirection to illustrate that they enter into the response in different ways,but there is no reason one might not have used a vector of variable magnitude,such as . Then theexpression would be simply . The form ofthis expression is an epicycloid that resembles a cardioid.
If we plot the responses for a series of accelerationsranging from no movement to a movement that completely suppresses thevestibular afferents activity. Then the response curves look like the followingfigure. The red curve is withminimal movement and the blue curve is with sufficient acceleration tocompletely suppress the activity when it is in the direction opposite thedirection of optimal responsiveness.
The responses to movements of an ampulla will be maximalwhen it is in the direction of , minimal when in the opposite direction and nil when in theplane perpendicular to the axis. So, each canal has a response surface that is like those drawn above,but rotated about the horizontal axis. If we assume similar response curves for each canal, then we can readoff the magnitudes of the responses in the different canals by computing theangle between the acceleration vector and the canal axis and plotting theradial vector at that angle. Wherethe vector intersects the surface of the appropriate acceleration the length ofthe vector will be the magnitude of the response. Because to the canals are approximately perpendicular,rotations about the axis of one canal will give a maximal response for thatcanal a near minimal response for its opposite on the other side of the headand near baseline responses in the other canals. The near baseline responses will be antagonistic in thecentral brainstem circuits therefore will approximately cancel each other whilethe maximal and minimal responses will show a near maximal difference.