Manual counting and measuring in microscopic images is very time consuming. Therefore, we develop automatic image analysis methods for microscopy. The images can be two or three dimensional and the sources can be light, fluorescence, or electron microscopes. The objects to measure can be cancer cells, proteins, or wood fibres and the purpose diagnostics or quality control.
New microscopy techniques are continuously developed, resulting in rapid acquisition of large amounts of 2D, 3D, time-series and multi-spectral data. Manual analysis and interpretation of such data is very time-consuming and subjective. Automatic extraction of quantitative measures using digital image processing and analysis is therefore of key importance.
Some features are not possible to quantify without the aid of a computer. We develop and adapt image processing and analysis algorithms for a large variety of microscopy-based applications, including wood fiber quality control, tissue screening, drug development, diagnostics, and more general research purposes. We develop methods for automated image acquisition and evaluation of sample quality and staining techniques, and our main focus is on preprocessing and normalization, segmentation, descriptive feature extraction, classification and visualization of images and results.
The image is made with fluorescence microscopy and shows neurons (green) colcultured with astorcytes (red) where all cell nuclei are blue. The mechanism of action of potential drugs can be studied by image-based quantification of phenotypic changes of the cells.
Electron microscopy allows for studying the shape and morphology of sub-cellular structures such as proteins and viruses at the nm level. Our current 2D research focuses on developing a miniaturized electron microscope, automated content driven image acquisition, and segmentation and texture measurements for virus diagnostics and nanoparticle analysis. We also develop interactive visualization and analysis methods for 3D tomographic data by studying the structure and flexibility of proteins and macromolecules in solution and in situ.
Transmission electron microscopy image of a fecal gastroenteritis patient sample containing Adenovirus.
Brightfield microscopy is used for tissue analysis in histopathology. We evaluate novel stains, develop color decomposition and segmentation methods, and design characteristic feature measures for automatic and quantitative cancer screening of prostate, breast, and cervix. In addition, we develop methods for reproductive toxicology research through morphological analysis of the male reproductive tract and improve therapeutic decision-making by quantitative measurements on the endothelial cells of the cornea.
Bright field image of a histology stained section of a testis. Automated image analysis has been applied to delineate the seminiferous tubule borders (black) and lumen (red).
Fluorescence microscopy enables imaging of fluorescent markers for identification of sub-cellular structures such as protein complexes, chromosomes, RNA, and genes relevant for biomedical research and diagnosis. Our current and previous research has focused on combining various types of image information, such as intensity, shape, and local gradients (in 2D and 3D) to mimic visual identification of sub-cellular structures. Image based screening, so called high-content analysis (HCA), of cultured cells or patient samples is used when searching for phenotypic changes related to chemical or genetic treatment. It provides an efficient tool for uncovering new biological pathways relevant to human disease. Screening also leads to the discovery of new probes for efficient and in-depth patient diagnosis, facilitating the selection and personalization of treatment. We develop methods for quantitative evaluation of phenotypic variations in cells and tissue, and over time. We have also approached HCA on model organisms, such as the worm C. elegans and Zebrafish, and develop specific algorithms for optical tomography, image segmentation and feature extraction.
Image shows optical tomography of a 5-day old zebrafish embryo after reconstruction and segmentation of the jaw bones.
UV light microscopy can visualize the lignin distribution in wood fiber walls and therewith quantify the quality of wood fibers. We develop automatic methods that detect and delineate softwood fibers and identify compression wood, which is present in almost every softwood tree harvested. Compression wood fibers have different mechanical properties and therefore they are considered detrimental for both construction wood, and pulp and paper purposes. We also use other imaging modalities, such as µCT, to study the distribution of fibers, fiber flocs and the void areas between fibers in different types of paper samples.
Petter Ranefall, PhD
Bioinformatician with focus on research support on image analysis
Centre for Image Analysis and Science for Life Laboratory
Quantitative Microscopy Group - Carolina Wählby