Interactive Mining for Large-scale Neuro-Morphological Datasets
1 online resource (96 pages) : PDF
University of North Carolina at Charlotte
In this dissertation, we aim to investigate advanced methods for the computational analytics of large-scale neuro-morphological datasets, which can help neuroscientists interactively explore neurons in real-time. Particularly, we tackle the neuro-morphological analytics into three inter-related components: 1) quantitative descriptions for 3D neuron morphologies, i.e., computing effective features that can differentiate subtle difference among massive neurons; 2) large-scale neuron mining, i.e., efficiently indexing neurons with similar morphologies in large-scale datasets; 3) interactively neuron exploration and visualization, i.e., developing neuron visualization tools and bring human in the loop to explore neurons in an interactive and immersive manner. We propose a series of methods in tackling problems related to the above three components. Regarding the quantitative description, we develop a deep learning framework based on neuron projection, which can transform 3D neurons into 2D images and learn effective neuron features. Regarding the large-scale neuron mining, binary coding methods are introduced, which can transform feature vectors into short binary codes for real-time indexing and mining. Regarding the interactive neuron exploration, we visualize 3D neurons using augmented reality (AR) techniques, where users can provide relevance feedback to further improve the mining performance. The proposed methods are validated on the currently largest neuron database including more than 58,000 neurons, achieving state-of-the-art performance in comparison with other related methods. More importantly, we demonstrate use cases of our framework in multiple neuron analysis and exploration tasks, showing its potential benefits in facilitating the research of neuroscience.
AUGMENTED REALITYDEEP LEARNINGLARGE-SCALE RETRIEVALNEURON MORPHOLOGYUSER INTERACTION
Zhang, ShaotingShi, XinghuaLu, AidongShin, Min
Thesis (Ph.D.)--University of North Carolina at Charlotte, 2018.
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.
Copyright is held by the author unless otherwise indicated.