Quantitative image analysis methods for applications in biomedical microscopy

This thesis describes a systematic procedure for performing quantitative analysis of biomedical microscopy images. The procedure is investigated with four case studies. The goal of the thesis is to illustrate the possibilities of automated image analysis in biomedical microscopy in general and to provide solutions for specific cases. The image analysis methods as well as the biological background of the case studies are presented in such a way that the thesis is accessible for researchers with a background in biology or in engineering. Quantitative information on biomedical samples has been traditionally obtained by visual inspection of the sample under the microscope. However, manual analysis is subjective, labor-intensive, and slow. All these problems can be overcome by developing automated image analysis methods. The presented procedure has six steps: sample preparation and image acquisition, image pre-processing, image segmentation, feature extraction, validation, and data analysis. The main emphasis of the thesis is in the three steps following sample preparation and image acquisition. They form the core of the procedure: we start with an image, essentially a matrix of real or integer values, and end up with a description of the image at a higher level of abstraction. A simple example would be the extraction of the number of cells in an image. Cell-specific spatial, spectral, and temporal features may be extracted as well. Validation of the obtained results and of the methods used is also very important. In the example given above, we want to be sure that the extracted number of cells is correct. The traditional way of validating results has been to compare the obtained results with results obtained by manual inspection of the cells under the microscope. The thesis proposes new simulation-based ideas for the validation step. The data analysis step uses the quantitative data obtained in the feature extraction step to obtain biologically relevant results. The first case study considers quantification of in vitro angiogenesis. The aims are two-fold. On the one hand, the developed quantitative image analysis method is used together with discrete growth models to study the global properties of angiogenesis. On the other hand, the method is used to study the effects of different agents on in vitro angiogenesis. The obtained information can be used in drug discovery, because one of the strategies for treating cancer is to inhibit angiogenesis. In the second case study the motivation is similar: a method for quantifying the effects of different agents on the invasive capability of cancer cells in an in vitro invasion assay is presented. The third case study considers quantification of budding yeast. There are two aims in this case study. The first aim is to obtain information on the cell cycle phase of the population by quantifying bud sizes. The second aim is to study peroxisome biogenesis in budding yeast. In the fourth case study an image analysis method is developed for quantifying lymph nodes in surgical specimens of colorectal cancer. A correlation between patient survival and the size and number of lymph nodes is found.