Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Thesis
Comparison of NCD and CNN based mitotic classification of neural stem cells in phase contrast microscopy
Master of Science (M.S.), Drexel University
Sep 2016
DOI:
https://doi.org/10.17918/etd-7127
Abstract
Automated mitosis detection is a major difficulty in segmentation and tracking algorithms. This thesis explores the implementation of an automated mitotic detection algorithm for phase-contrast data into a segmentation and tracking platform called LEVER2. We implement two separate classification algorithms - one based on an estimate of the pairwise normalized compression distance (NCD) and one deep learning implementation using convolutional neural networks (CNN) - in combination with local statistics in order to limit the search region for mitotic events. The CNN provided a significantly higher detection rate while the NCD provided a significantly lower false positive rate. The NCD classifier generated an overall detection rate of 0.736 and an overall false positive rate of 0.0083, while the CNN approach generated an overall detection rate of 0.880 and a false positive rate of 0.0567. However, it was also determined that using a bi-modal classifier incorporating the peak normalized cross-correlation of the image with the t - 1 frame the CNN implementation could outperform the NCD classifier - achieving a false alarm rate of 0.0079 with a detection rate of 0.809 by thresholding out cell images that were above a threshold peak cross-correlation.
Metrics
29 File views/ downloads
22 Record Views
Details
- Title
- Comparison of NCD and CNN based mitotic classification of neural stem cells in phase contrast microscopy
- Creators
- Michael Marino - DU
- Contributors
- Andrew R. Cohen (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Master of Science (M.S.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- viii, 59 pages
- Resource Type
- Thesis
- Language
- English
- Academic Unit
- College of Engineering (1970-2026); Electrical (and Computer) Engineering [Historical]; Drexel University
- Other Identifier
- 7127; 991014632233904721