東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

A Bayesian Approach to Spike Sorting...

Greene, Patrick.

FindBook

Google Book

Amazon

博客來

A Bayesian Approach to Spike Sorting of Neural Data via Source Localization.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	A Bayesian Approach to Spike Sorting of Neural Data via Source Localization./
作者:	Greene, Patrick.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
面頁冊數:	117 p.
附註:	Source: Dissertations Abstracts International, Volume: 80-01, Section: B.
Contained By:	Dissertations Abstracts International80-01B.
標題:	Neurosciences. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10831343
ISBN:	9780438118256

A Bayesian Approach to Spike Sorting of Neural Data via Source Localization.
Greene, Patrick.

A Bayesian Approach to Spike Sorting of Neural Data via Source Localization. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 117 p.

Source: Dissertations Abstracts International, Volume: 80-01, Section: B.

Thesis (Ph.D.)--The University of Arizona, 2018.

This item must not be sold to any third party vendors.

This dissertation describes a novel mathematical algorithm for extracting spike data and positional information from extracellular electrophysiological neural recordings. By capturing the electrical signals emitted by individual neurons using a thin, conducting probe inserted into the brain of an animal, such recordings allow us to understand how groups of neurons process and encode information about the animal and its environment, and are one of the basic tools of modern neuroscience. However, these recordings generally cannot be used directly because factors such as background activity, movement, and electrical fluctuations produce significant amounts of noise in the recorded data. In addition, nearby cells often code for different features, so simply averaging over the data would result in loss of information. Therefore, as an essential first step, one must extract the underlying signals (spikes) contained within the raw data, and assign these spikes to particular neurons. This process is called spike detection and sorting, and the degree of accuracy to which it can be done directly affects the quality and reliability of all downstream analysis. This dissertation consists of three main components. The primary component is the spike sorting algorithm itself, whose overall mathematical framework is described in chapters 3 and 4. There are two key ideas. The first is the usage of a dipole-based generative model for recorded waveforms. This dipole approximation of a neuron, which captures the empirical falloff in signal strength with distance from the probe, allows us to estimate the position from which signals originate. This in turn helps us estimate waveform shapes and the number of neurons being recorded from. The second key idea is to incorporate this dipole model within an extended Bayesian version of a gaussian mixture model. This gives us a principled way to deal with many of the issues that arise in spike sorting which cannot be resolved by previous methods. We then implement this mathematical framework in the python programming language and test it on both simulated and experimental data. We compare it against a basic mixture model approach and find that it does indeed improve accuracy. The second component of this dissertation is a relatively realistic model of the extracellular signal generation and recording process (the "forward model") which we describe in chapter 2. We construct this model in order to better understand the physics of extracellular signals, and how they are affected by probe position and neuron geometry. Experimentation with this model allowed us to formulate the simplified physical model which was later incorporated into our final spike sorting algorithm. The forward model also allows us to generate realistic test data which we use to judge the accuracy of our method. The final component of this dissertation is an investigation into the effect of probe geometry and the dipole prior distribution on how well we can estimate neuron positions. This makes extensive use of the forward model described above, and is the content of chapter 5. We show that the algorithm we have developed is robust across a range of prior sizes, and determine the probe geometry which produces optimal localization accuracy.

ISBN: 9780438118256Subjects--Topical Terms:

588700
Neurosciences.

A Bayesian Approach to Spike Sorting of Neural Data via Source Localization.
LDR:04385nmm a2200337 4500 001 2210446
005 20191121124222.5
008 201008s2018 ||||||||||||||||| ||eng d
020 $a 9780438118256
035 $a (MiAaPQ)AAI10831343
035 $a (MiAaPQ)arizona:16444
035 $a AAI10831343
040 $a MiAaPQ $c MiAaPQ
100 1 $a Greene, Patrick. $3 3437585
245 1 0 $a A Bayesian Approach to Spike Sorting of Neural Data via Source Localization.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 117 p.
500 $a Source: Dissertations Abstracts International, Volume: 80-01, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Lin, Kevin K.
502 $a Thesis (Ph.D.)--The University of Arizona, 2018.
506 $a This item must not be sold to any third party vendors.
520 $a This dissertation describes a novel mathematical algorithm for extracting spike data and positional information from extracellular electrophysiological neural recordings. By capturing the electrical signals emitted by individual neurons using a thin, conducting probe inserted into the brain of an animal, such recordings allow us to understand how groups of neurons process and encode information about the animal and its environment, and are one of the basic tools of modern neuroscience. However, these recordings generally cannot be used directly because factors such as background activity, movement, and electrical fluctuations produce significant amounts of noise in the recorded data. In addition, nearby cells often code for different features, so simply averaging over the data would result in loss of information. Therefore, as an essential first step, one must extract the underlying signals (spikes) contained within the raw data, and assign these spikes to particular neurons. This process is called spike detection and sorting, and the degree of accuracy to which it can be done directly affects the quality and reliability of all downstream analysis. This dissertation consists of three main components. The primary component is the spike sorting algorithm itself, whose overall mathematical framework is described in chapters 3 and 4. There are two key ideas. The first is the usage of a dipole-based generative model for recorded waveforms. This dipole approximation of a neuron, which captures the empirical falloff in signal strength with distance from the probe, allows us to estimate the position from which signals originate. This in turn helps us estimate waveform shapes and the number of neurons being recorded from. The second key idea is to incorporate this dipole model within an extended Bayesian version of a gaussian mixture model. This gives us a principled way to deal with many of the issues that arise in spike sorting which cannot be resolved by previous methods. We then implement this mathematical framework in the python programming language and test it on both simulated and experimental data. We compare it against a basic mixture model approach and find that it does indeed improve accuracy. The second component of this dissertation is a relatively realistic model of the extracellular signal generation and recording process (the "forward model") which we describe in chapter 2. We construct this model in order to better understand the physics of extracellular signals, and how they are affected by probe position and neuron geometry. Experimentation with this model allowed us to formulate the simplified physical model which was later incorporated into our final spike sorting algorithm. The forward model also allows us to generate realistic test data which we use to judge the accuracy of our method. The final component of this dissertation is an investigation into the effect of probe geometry and the dipole prior distribution on how well we can estimate neuron positions. This makes extensive use of the forward model described above, and is the content of chapter 5. We show that the algorithm we have developed is robust across a range of prior sizes, and determine the probe geometry which produces optimal localization accuracy.
590 $a School code: 0009.
650 4 $a Neurosciences. $3 588700
650 4 $a Applied Mathematics. $3 1669109
650 4 $a Statistics. $3 517247
690 $a 0317
690 $a 0364
690 $a 0463
710 2 $a The University of Arizona. $b Applied Mathematics. $3 1266624
773 0 $t Dissertations Abstracts International $g 80-01B.
790 $a 0009
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10831343