Image acquisition processing and machine vision moving from images to data files

We considered the type of "objects" we could create during the presentation stage and how to optimally image them as a single problem to be solved. This tact enabled us to critically engineer features for data acquisition in robust and operationally simple ways. As such, we developed a high-speed imaging instrument (named, "genome Zephyr''), or a single molecule "scanner," built around a standard fluorescence microscope featuring full computer control over focusing, sample positioning and digital camera functionalities, thus enabling user-free operation launched from a friendly interface. Throughput is greatly enhanced by a distributed laser illumination system offering stable, bright, monochromatic illumination to all of the Zephyr scanners in our laboratory. Essentially, the Zephyr automatically acquires strings of contiguous, overlapping micrographs, by tracking stripes of deposited DNA molecules laid down by the microfluidic system. Automatic image processing takes these images, corrects for uneven illumination, and then overlaps them (~ 100-150 images) maintaining proper registration into one "superimage" potentiating downstream machine vision approaches. Since we are using high-power fluorescence objectives, we realize a spatial resolution of about 300 nm. All of these features synergize critical machine vision steps that automatically identify molecules, discern "daughter" restriction fragments, and finally estimate their sizes (in kilobases), based on integrated fluorescence intensity measurements. This last step creates a restriction map for each imaged molecule, and an analyzed surface yields thousands of single molecule maps or barcodes as compact data files.

0 0

Post a comment