General-Purpose High-Throughput Image-Processing-Ingestion Pipeline

Back to [Bio]nanotechnology & Programmable Matter

Partner: Wyss Institute at Harvard University

We are working on a general-purpose high-throughput image-processing-ingestion pipeline, taking advantage of the high processing power of todays multicore machines and higher memory bandwidth. The goal is to collect images from an arbitrary number of sources and execute multi-stage image processing on them, including filtering, object recognition, alignment, and data extraction. The initial application of this pipeline is DNA sequencing, but the pipeline is designed to be general purpose, with a skew towards biological imaging applications.

Description of the images

These are aligned composite images of the a single typical flowcell field of single strand rolonies, or clews of DNA tagged with 4 different fluorophores per base A,C,G,T. The images are acquired on a high bit-depth greyscale camera and then down sampled to 8 bits to allow compositing, imaging only one base at at time, fluorescing at on wavelength. Since color digital images typically only have three 8-bit color channels, the first image composites all four images, using yellow as a pseudo-fourth color showing all rolonies in the field, and the second image, a little easier on the eyes, composites only three of the images with three different colors, using only red, green, and blue.