Image Segmentation, Processing and Analysis in Microscopy and Life Science

Microscopes have been used for more than 400 years to understand biological and biomedical processes by visual observation. Science is the art of observing, but science also requires measuring, or quantifying, what is observed. Research based on microscopy image data therefore calls for methods for quantitative, unbiased, and reproducible extraction of meaningful measurements describing what is observed. Digital image processing and analysis is based on mathematical models of the information contained in image data, and allows for automated extraction of quantitative measurements. Automated methods are reproducible and, if applied consistently and accurately across experiments with positive as well as negative controls, also unbiased. Digital image processing is further motivated by the development of scanning microscopes and digital cameras that can capture image data in multiple spatial-, time-, and spectral-dimensions, making visual assessment cumbersome or even impossible due to the complexity and size of the collected data.

The process of analyzing a digital image is usually divided into several steps, where the objects of interest are first identified, or ‘segmented’, followed by extraction of measurements and statistical analysis. This chapter starts from the basics of describing images as matrices of pixel intensities. Emphasis is thereafter put on image segmentation, which is often the most crucial and complicated step. A number of common mathematical models used in digital image processing of microscopy images from biomedical experiments are presented, followed by a brief description of large-scale image-based biomedical screening.