Information Content Analysis in Automated Fluorescence Microscopy
| Authors | |
|---|---|
| Year of publication | 2010 |
| MU Faculty or unit | |
| Citation | |
| Description | This thesis deals with automated fluorescence microscopy, in particular with automatic content analysis algorithms, which is a concept that we introduced as a generalization of the well-established autofocusing procedure. The aim of such algorithms is to enable automatic selection of regions of interest to be acquired and subsequent acquisition of images with high information content (i.e. containing objects of interest), in our case in fluorescence mode. The emphasis is on accuracy and reliability. Speed is also a critical factor due to the common phenomenon of sample photobleaching. The main challenge to the state of the art methods is the existence of more planes (more focus positions) with high information content to be acquired, which is the consequence of specimen thickness or simply of the fact that all the imaged objects do not lie in one plane. For the wide-field mode (as compared to the confocal mode) there is even a lack of a suitable function that would measure the image content, regardless of the number of planes with high content. In all cases, none of the previous approaches was able to provide satisfactory results. These issues are identified in this work and demonstrated on a set of tests. Several solutions to the detected problems and enhancements of the previous procedures are provided in this work. Firstly, for the wide-field mode, a function for measuring information content is suggested. Secondly, a new technique based on field of view division is developed for detecting planes with high information content that often remain undetected by naked eye. Thirdly, an adaptive algorithm is developed for speeding up the detection of multiple planes with high information content. The work provides also many recommendations for the selection and tuning of proper algorithms and parameters in various conditions. The proposed methods outperform the previously used approaches on given test images and are thus very promising for practical use. |
| Related projects: |
|