2016 Fiscal Year Annual Research Report
New Data-Driven Tools to Quantify Heterogeneous Microenvironments in Live Cell Images
Project/Area Number |
15K18511
|
Research Institution | Hokkaido University |
Principal Investigator |
Taylor Nicholas 北海道大学, 電子科学研究所, 特任助教 (50750824)
|
Project Period (FY) |
2015-04-01 – 2017-03-31
|
Keywords | Raman Microscopy / Live-Cell Imaging / Data Science / Machine Learning / Statistics |
Outline of Annual Research Achievements |
The inherent weakness of the Raman scattering signal along with the presence of various forms of contamination and noise greatly hinders objective quantification of cellular detail. A statistical method to quantify background contamination was developed and and tested. The method performs well for sparsely distributed samples, but its efficacy decreases as signal density increases. A new denoising scheme using the statistical behavior within the images was developed. Preliminary tests outperform traditional denoising methods with the advantage of having no adjustable parameters. Even with the inclusion of such enhancements, accurate characterization of cellular spectra remains challenging. Supervised alternatives to traditional unsupervised clustering methods are currently being explored.
|
Research Products
(2 results)