With the advent of deep convolutional neural networks, conventional machine learning algorithms have been increasingly displaced, particularly in the computer vision domain. Notably, convolutional neural networks have proven indispensable in image classification, segmentation, and object detection which has driven innovation in self-driving cars, facial recognition, and pedestrian detection to name a few. Similarly, the successful application of DCNNs to medical imaging could expedite experimentation and analysis that may lead to future treatments or contribute to improvements in the standard of care. However, the scarcity and uniqueness of medical imaging pose challenging scenarios that require novel methodologies. Likewise, time-series classification tasks are also hampered by limited and unique data. Still, the application of DCNNs to this task also outperforms conventional methods and demonstrates the flexibility of CNNs in automatically extracting features in various situations. In this thesis, three distinct tasks with limited data are addressed using DCNNs.
First, object localization is applied to a 3 dimensional MRI dataset to facilitate cochleae extraction. Experimental results show that localization is possible within a 1.71 voxel error. Second, this thesis explores CNN architectures for time-series classification of shark behaviors based on data collected by the CSULB Shark Lab. Experimental results show a weighted-average F1-score of 80.0%. Lastly, synaptic cleft segmentation is performed on serial section Transmission Electron Microscopy volumes, where submission to the CREMI challenge ranks 1st on the leaderboard.
|Commitee:||Fu, Bo, Kim, Eun Heui|
|School:||California State University, Long Beach|
|Department:||Computer Engineering and Computer Science|
|School Location:||United States -- California|
|Source:||MAI 81/7(E), Masters Abstracts International|
|Subjects:||Computer science, Artificial intelligence|
|Keywords:||Biomedical, Classification, CREMI, Deep learning, Object detection, Segmentation|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be