Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

We describe how spectral imaging, linear un-mixing and cluster computing have been combined to aid biomedical researchers and allow the spatial segmentation and quantitative analysis of immunohistochemically stained tissue section images. A novel cost-effective spectral imager, with a bandwidth of 15 nm between 400 and 700 nm, allows us to record both spatial and spectral data from absorptive and fluorescent chemical probes. The linear un-mixing of this data separates the stain distributions revealing areas of co-localisation and extracts quantitative values of optical density. This has been achieved at the single-pixel level of an image by non-negative least squares fitting. This process can be computationally expensive but great processing speed increases have been achieved through the use of cluster computing. We describe how several personal computers, running Microsoft WindowsXP, can be used in parallel, linked by the MPI (Message Passing Interface) standard We describe how the free MPICH libraries have been incorporated into our spectral imaging application under the C language and how this has been extended to support features of MPI2 via the commercial WMPI II libraries. A cluster of 8 processors, in 4 dual-Athlon-2600+ computers, offered a speed up of a factor of 5 compared to a singleton. This includes the time required to transfer the data throughout the cluster and reflects a processing efficiency of 0.62 (a Cluster Efficacy of 3.0). The cluster was based on a 1000Base-T Ethernet network and appears to be scalable efficiently beyond 8 processors.

Original publication




Conference paper

Publication Date





53 - 61