Analysis of spectrally resolved autofluorescence images by support vector machines

A. Mateasik; D. Chorvat; A. Chorvatova

doi:10.1117/12.2001371

Translator Disclaimer

22 February 2013 Analysis of spectrally resolved autofluorescence images by support vector machines

A. Mateasik, D. Chorvat, A. Chorvatova

Author Affiliations +

Proceedings Volume 8588, Multiphoton Microscopy in the Biomedical Sciences XIII; 85882J (2013) https://doi.org/10.1117/12.2001371
Event: SPIE BiOS, 2013, San Francisco, California, United States

Abstract

Spectral analysis of the autofluorescence images of isolated cardiac cells was performed to evaluate and to classify the metabolic state of the cells in respect to the responses to metabolic modulators. The classification was done using machine learning approach based on support vector machine with the set of the automatically calculated features from recorded spectral profile of spectral autofluorescence images. This classification method was compared with the classical approach where the individual spectral components contributing to cell autofluorescence were estimated by spectral analysis, namely by blind source separation using non-negative matrix factorization. Comparison of both methods showed that machine learning can effectively classify the spectrally resolved autofluorescence images without the need of detailed knowledge about the sources of autofluorescence and their spectral properties.

Citation Download Citation

A. Mateasik, D. Chorvat, and A. Chorvatova "Analysis of spectrally resolved autofluorescence images by support vector machines", Proc. SPIE 8588, Multiphoton Microscopy in the Biomedical Sciences XIII, 85882J (22 February 2013); https://doi.org/10.1117/12.2001371

ACCESS THE FULL ARTICLE