Compact high-speed panoramic visual arena and custom signal processing pipeline applied to complete mapping of insect visual receptive fields
File(s)
Author(s)
Ko, Daniel
Type
Thesis
Abstract
Vision is the primary sensory modality for many animals, and efficient exploitation of its rich information content allows impressive feats of visuomotor coordination in scenarios such as prey interceptions. Studying the neural basis that governs these behaviours typically requires three steps: 1) controlling the visual input of the animal; 2) recording responses from appropriate neural substrate; 3) analysing these responses with respect to the visual input. Careful choice of hardware and software tools in each of these is paramount to the success of the study. My PhD project aimed to solve several technological hurdles for insect visual neuroscience using the dragonfly as a model.
Among visual animals, insects often have relatively poor spatial acuity but faster vision. This has led to visual stimulation methods in insect neuroscience to prioritise speed over resolution. However, some outliers such as dragonflies require visual stimuli with both high speed and resolution. In response to this, I designed and implemented a high-performance visual stimulation system which is detailed in Chapter 2. Furthermore, due to the nature of insect neural signals, existing spike sorting tools designed for vertebrate recordings perform poorly with insect models. In Chapter 3, I highlight these performance issues and present a new GUI-based semi-automatic spike sorter I developed specifically for high-accuracy sorting of neurons with low firing rates.
To test and validate my designs, I applied them in the study of dragonfly target detection and tracking. Target Selective Descending Neurons (TSDNs) have been identified as strong candidates for the neural basis of the transformation of target information to wing commands. However, the boundaries of their receptive fields and how they spatially interact for target detection are unknown. In Chapter 4, I uncover these receptive field boundaries in the dragonfly Sympetrum striolatum and discuss their implications in the encoding of target information.
Among visual animals, insects often have relatively poor spatial acuity but faster vision. This has led to visual stimulation methods in insect neuroscience to prioritise speed over resolution. However, some outliers such as dragonflies require visual stimuli with both high speed and resolution. In response to this, I designed and implemented a high-performance visual stimulation system which is detailed in Chapter 2. Furthermore, due to the nature of insect neural signals, existing spike sorting tools designed for vertebrate recordings perform poorly with insect models. In Chapter 3, I highlight these performance issues and present a new GUI-based semi-automatic spike sorter I developed specifically for high-accuracy sorting of neurons with low firing rates.
To test and validate my designs, I applied them in the study of dragonfly target detection and tracking. Target Selective Descending Neurons (TSDNs) have been identified as strong candidates for the neural basis of the transformation of target information to wing commands. However, the boundaries of their receptive fields and how they spatially interact for target detection are unknown. In Chapter 4, I uncover these receptive field boundaries in the dragonfly Sympetrum striolatum and discuss their implications in the encoding of target information.
Version
Open Access
Date Issued
2023-12-22
Date Awarded
01/11/2024
License URL
Advisor
Lin, Huai-Ti
Sponsor
European Research Council
Publisher Department
Bioengineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)