Extracting Cancer pagurus stomatogastric ganglion pyloric rhythm frequency via voltage-sensitive dye imaging data using signal processing techniques

Abstract

Voltage-sensitive dye imaging (VSDI) has been widely used in the past few decades in both vertebrates and invertebrates to study, in vitro and in vivo, the nervous systems. Cancer pagurus is a seawater crab whose nervous system has a ganglion, the stomatogastric ganglion (STG) that contains a relatively small number of neurons and two rhythm forming central pattern generators (CPGs). The pyloric rhythm is one such spontaneous rhythm that can be readily observed in vitro, which makes the STG an ideal ganglion to study using VSDI. However, a major impediment to the effectiveness of VSDI is that the optically recorded data is often noisy with poor signal-to-noise ratios (SNR), rendering it difficult to study and analyse.
This thesis describes the first-ever development of computational signal processing procedures that sought to extract the pyloric rhythm directly from the VSDI data, thus facilitating an accurate identification of the individual neurons in the pyloric circuit. Specifically, a multiresolution procedure based on the sequential Singular Spectrum Analysis (s-SSA) was first constructed to separate the pyloric rhythm from the noisy VSDI recording, enabling potential pyloric neurons to be detected by the presence of the pyloric frequency in the computed spectra of the respective cells. To facilitate identifying the pyloric neurons, the duty cycle (DC) was devised as a biometric, and the corresponding ratio of harmonics (RH) was determined in terms of the harmonic content of the spectrum computed for each cell/neuron as described above. Here, the instantaneous phase of the detected pyloric rhythm was also estimated, allowing it to be compared and aligned with the three distinctive pyloric phases (PD-, LP- and PY-timed) readily measured on the lateral ventricular nerve (lvn). As proof of concepts, finally, an automated method to determine the pyloric frequency directly from VSDI data was developed, over a range of SNRs, demonstrating the possibility to identify prospective pyloric neurons based on the estimated DCs and respective phase shifts measured against the analogue lvn recording.