|Comment||Vascular Biology group, Department of Genetics and Pathology, UU|
|Title||A Random Walk in Image Analysis|
|Abstract||In addition to working with microscopes, mostly fluorescence, I used and developed image analysis software for around twenty years, based on a Semper6 kernel (Synoptics, Cambridge, UK). My earliest work was on methods for generating noise free perimeter fractal dimensions. We developed a routine for converting a 2D perimeter into a sequential list of pixels and then reducing resolution by smoothing the sequence over an increasing number of neighbours (Part. Part. Systems Characterisation, 1994, 418-425, 11). Porous materials - using simulated diffusion to examine porous 2D structures by using flux as a measure of structure. Reaching a steady state is computationally demanding, an alternative is to calculate the resistance to flow using an aggregate of the widths of every channel measured along the midline – like a network of electrical resistors. Simulated particle diffusion over non flat surfaces (Nature Methods. 2010, 7,170-1). Measuring colocalization using correlation - we have developed method using two sets of paired images that is independent of image quality (J. Microscopy, 2008, 230, 121-133).
In the Vascular Biology group, we are working on software to
1) Estimate the fraction of mouse retina covered by sprouting blood vessels – we use semi automated stereological approach as the difference between sprouts and normal vessels is not easy to define precisely enough for automation.
2) Concentration gradients along blood vessel sprouts composed of a series of joined cells. We want to discover if various cellular markers change in abundance from the tip of a sprout. This is achieved by defining slicing the sprout into ROIs a different distances from the tip and measuring the intensity in each ROI. The ROIs are created from equidistant points along the midline of the sprout starting from the tip.
3) To automatically count and measure projections (small sprouts) from cells. Possibly by defining a sprout by its diameter and tips or ends by their high curvature.
4) We are also interested in characterizing vascular networks from 3D data sets when the vessels are delineated using markers of the vessel wall or the lumen. Characterization could be partially achieved using methods for porous material mentioned above if they can be extended to 3D.
5) Distribution of fluorophores and particles relative to cellular structures – we use distance transforms in 2D but need to extend the measurements to 3D.