Using Video Microscopy to Characterize Micromechanics of Biological and Man-Made Micromachines (invited)

Dennis M. Freeman and C. Quentin Davis

Presented at the Solid-State Sensor and Actuator Workshop

Hilton Head Island, SC, June 1996.

Part III: Application to Hearing

still picture (155K gif)
These two images and the associated video illustrate the motions of the tectorial membrane (left) and hair bundles (right) that result from a moderately loud (89 dB SPL) sound source. The 513 Hz frequency was chosen to match the most sensitive frequency for these hair cells.
still picture (12K gif)
Keep in mind that our goal is to determine how different structures in the inner ear interact, and to test modern theories of cochlear mechanics. Specifically, our measurements allow us to test the idea that the tectorial membrane is a resonant structure.
still picture (155K gif)

These images are the first images of the motions of hair bundles and their overlying tectorial membrane. They are the first measurements that allow direct tests of previous theories of cochlear mechanics. These images clearly indicate that the tectorial membrane is NOT moving more than the hair bundles. The tectorial membrane is in fact moving much less. Thus these measurements do not support the notion that the tectorial membrane is a resonant structure. This result should be regarded as preliminary until sufficiently many control experiments have been done to eliminate the possibility of experimental error. However, to date, we have repeated this experiment with 12 different lizard ears, and we have never seen resonance of the tectorial membrane.

still picture (113K gif)
We can get much more from this same data set. Specifically, we can electronically zoom in to look at mechanics at the level of individual cells.
still picture (74K gif)
The left part of this slide shows a schematic drawing of a hair bundle and overlying tectorial membrane. The right panels show images from different planes of section that have been stacked up to produce a 3D image of a hair bundle. By showing such 3D images as a function of time, the associated video can convey a qualitative idea of the 3D motions of a hair bundle.
still picture (75K gif)
Now we can apply our computer vision algorithms to obtain quantitative motion estimates. This slide and the associated video shows motion estimates for each of the six panels in the previous slide. The waveforms show the displacements of the associated images as a function of time. As we saw previously, motion of the tectorial membrane is smaller than that of the hair bundles. Motion of the hair bundles is about 0.5 micrometers peak-to-peak and that of the top of the tectorial membrane is about 0.18 micrometers. However, the quantitative estimates also reveal a systematic change in phase. The top of the tectorial membrane lags the base of the hair cell by nearly 90 degrees. This change of phase through the tectorial membrane has not previously been suggested. It represents a new mode of motion: a mode that hadn't been conceived before these measurements.
still picture (77K gif)
The quantitative motion estimates allow other interesting manipulations. For example, we can shift each of the images into compensate for the motion at the base of the hair bundle. The resulting images and associated video show motions relative to the base, i.e., in a frame of reference attached to the hair cell. When we estimate motions of these images with our computer vision algorithms, the motions at the base are greatly reduced. Now we can use the peak-to-peak motions near the tip of the hair bundle to estimate the angular displacement of the hair bundle. The plane near the tip is 6 micrometers above the plane near the base, and the motions near the tip are about 0.3 micrometers peak-to-peak. Therefore, the angular displacement is about 3 degrees peak-to-peak.
still picture (64K gif)
We can zoom in even more to investigate the motions between sensory hairs within a bundle.
still picture (81K gif)
This slide and the associated video shows a section through the center part of the hair bundle. Translations of the images result both because the base is translating and because the bundle is rotating. In addition to the translation, there also appears to be a modulation of the distance between the individual sensory hairs.
We can use our motion estimation algorithm to estimate the motion of the right edge of the hair bundle and then shift the images to compensate for that motion. The resulting images and associated video show relative motions between sensory hairs. For example, the distance between the hairs near the center of the bundle changes as the bundle translates and rotates. Changes in distance between sensory hairs are important because they affect the stress in tip links.
Previously, it was thought that sensory hairs remain parallel as the bundle rotates, as was shown in the previous animation. This conception was based on experiments in which bundles were displaced with tiny glass fibers or with a water jet and observed with a microscope. In hindsight, it is not difficult to imagine how results are different for dynamic stimuli. The tip links provide elastic connections between hairs. But the surrounding fluid provides viscous drag. Thus it is not surprising that motions of shorter hairs lag those of longer hairs. But this relation was not suspected previously.
still picture (51K gif)
In summary, we've applied computer microvision to gain insights into cochlear micromechanics at the level of single sensory hairs, at the level of hair bundles, and for the entire inner ear (see video).

Let me now move on to demonstrate the application of computer microvision to MEMS.

Application to MEMS