One of the main limitations to the performance of the electrode systems is the manually controlled positioning of the electrode in close proximity, the near-field, to the cell membrane: (1) The stability and resolution in probe placement depend on the operator skills in attending and promptly adjusting the location of the probe with respect to mobile specimens. From studying the near-field characteristics of the physiological signal in quantitative terms we conclude that the probe tip should be placed at sub-micron distances from the target and that the tip-target relationship should be held constant with a confidence below 100 nm. To date both requirements can hardly be fulfilled. (2) The present qualitative nature of manual control precludes the exploration of many potential applications of the instrumentation. As the probe-target geometry is not quantified throughout signal recording reliable spatial mapping and comparative studies of flux patterns from several samples are made difficult and sometimes impossible. In this project we plan to develop a machine vision system supporting the regulation of the probe. We will digitize image data delivered by a light microscope observing the scene and extract the information needed for probe control by using computer vision methods. Such a system will boost the capabilities of the available technology in the following aspects: (1) Both stability and resolution in positioning the electrode with respect to motile targets will be increased by at least an order of magnitude. Higher positional stability will result in a strong reduction of artifacts in the probe signal which are induced by uncertainties in probe-target geometry. Better resolution in probe control will allow us to place the probe closer to the target. (2) A new level of quantitative information derivable from the recorded data will be reached. It will become possible to continuously measure the distance between probe tip and cell boundary with high accuracy. This will enable us to spatially map transmembrane ion transport and to perform comparative studies between several preparations. (3) Less effort in performing the recordings and only minimal knowledge about the technology will b e necessary to obtain a data series of reasonable quality. To implement a controller for automated position control based on machine vision we have to implement (1) algorithms for the visual tracking of probe and membrane, (2) tools for the measurement of the distance between membrane and probe tip, and (3) a module which regulates the position in relation to the cell membrane. In the past 6 month we have studied different algorithms for the tracking of the relevant features in the scene. We have applied a pattern matching technique (Danuser et al. 1996) for the tracking of the probe tip. Basically, the algorithm models the image variability between consecutive frames due to motion of the probe. From the estimated parameters of the variability model one can derive the actual displacement of the probe between the frames. In several tests we have proven that the sensitivity of this method in tracking the probe tip reaches the level of single nanometers, which is at about the theoretical limit of what can be observed light optically (Inoue 1989, Danuser 1997). We experienced some problems with this algorithm when changing the contrast mode. As a next step we will apply a new mathematical formulation of the pattern variability in order to become algorithmically less sensitive with respect to the various contrast modes one may wish to use for optimal visualizatio n. Efforts in tracking the membrane as randomly evolving contour are reported in the article """"""""Computer Vision based Membrane Tracking"""""""", below. G. Danuser and E. Mazza 1996. Observing Deformations of 20 Nanometer with a Low Numerical Aperture Light Microscope. In Optical Inspection and Micromeasurements, European Optical Society, SPIE Vol. 2782:180-191. G. Danuser 1997. Quantitative Stereo Vision for the Stereo Light Microscope: An Attempt to Provide Control Feedback for a Nanorobot System. Diss. ETH No. 12191 (O. Kuebler, H. Tiziani, M. Stricker, advisors). ISBN 3-905588-00-5 S. Inoue 1989. Imaging of Unresolved Objects, Superresolution, and Precision of Distance Measurement with Video Microscopy. In Methods in Cell Biology, Vol. 30, Chapter 3, pp. 85-112.
Showing the most recent 10 out of 144 publications
