Improved Patch-Clamp Techniques Based On Inverse Problem Formulation

Patch-clamp techniques, the "gold standard" in ion channel recording, havedramatically advanced our knowledge about functionality of ion channels in transmittingneuronal signals, and become a indispensable validation method in drug discovery process.However, the problem of labor-and skill-intensive manipulation and low spatial resolutionlimit the application of this technique. To improve the automation and integration level ofthe patch-clamp amplifier, we introduced modern signal processing techniques, which wereexpected to make contribution to facilitating the calibration process and minimizing the sizeof the equipment.The basic idea of minimizing the patch-clamp amplifier is to replace the conventionalanalog signal processing circuits with digital signal processing program. The architecture ofthe improved electronic system would be greatly simplified to contain only the headstageand the data acquisition board. To do so, the functions of previous analog signal processing,e.g. frequency correction, should be achieve by running software.This thesis firstly reviews some typical signal processing problems in the patch-clampexperiment using inverse problem formulation. These problems are related to the threepathway (i.e., the main signal pathway, the capacitance compensation pathway, and theseries resistor compensation pathway) in the patch-clamp amplifier. The inverse problemformulation not only clears up the thoughts in the analog signal processing, but alsoprovides a basic mathematical model for the digital signal processing.This dissertation mainly focuses on solving the inverse problem of convolution, i.e. thedeconvolution, which is really a challenge. The general concept of solving this kind ofproblem is to convert the inverse problem into a forward problem and to find the optimalanswer iteratively. However, the iterative method suffers from some inherent disadvantages,such as no guaranteed convergence. Here, I employed two non-iterative techniques, whichwere cross-correlation technique and subspace system identification technique, based onARMA model and state space model respectively. With above two techniques, we characterized the specific pathway of the patch-clamp amplifier with a non-parameterizedor parameterized model, i.e. the convolution kernel and the state space model. Differentfrom conventional frequency spectrum, these two model provides transient properties of thepathway, which are very important in kinetic analysis.Given the kernel of the main signal pathway, I developed a novel non-iterativeautomatic fast capacitance compensation method—K-method. Using the subspace systemidentification technique, I accurately estimated the series resistance with a high value, anddeveloped a brand-new software-based high-frequency compensation method—SHB.These new methods were validated using a model circuit or real cells.The new methods not only open the door to fabricating a highly-integrated patch-clamparray with giga-ohm gain, but they also provide examples of successful combination withmodern signal processing theory for patch-clamp techniques.