Andrew Gait

Dr. Andrew Gait

Research Software Engineer
Office IT302
School of Computer Science,
IT Building,
University of Manchester,
Oxford Road,
Manchester
M13 9PL
 
Email: Andrew.Gait[at]manchester.ac.uk

SpiNNaker toroidal connections

Current work

Since February 2017 I have been part of the software team for the SpiNNaker project, led by Prof. Steve Furber, in the APT group at the School of Computer Science.

This involves day-to-day support of users of the software across various applications, mainly related to neuromorphic modelling. Our software stack can be found on GitHub, and we also maintain associated documentation (including installation instructions).

SpiNNaker is not necessarily limited to neuromorphic modelling: see my most recent poster from RSE 2018.


Previous work

I worked at the Wolfson Molecular Imaging Centre on various software solutions to problems in image analysis. In particular, I designed a software tool to investigate parallel coordinates across multiple co-registered image sets.

Parallel coordinate software with oxygen-enhanced MRI

An example using the software tool for parallel coordinates with oxygen-enhanced MRI.


Previous to that (2009-2014), I worked on the Research into OsteoArthritis in Manchester (ROAM) project. I work with Prof. Tim Cootes on image analysis of MRI scans of knees.

Sagittal image of a patellar bone marrow lesion

An image of the progression of a bone marrow lesion during treatment with a brace.

Mean knee image from group-wise registration method

A mean image using generalised group-wise registration of the knees of 17 patients.

Gadolinium contrast-enhanced image sequence

A Gadolinium-enhanced dynamic image sequence; images in this sequence are taken at a time resolution of roughly 23 seconds. The contrast agent is added at around 50-60 seconds and can be seen on the image first in the major artery, and later in the enhanced regions in the patellofemoral region towards the top of the image.

The uptake of the contrast agent can be modelled at each pixel on each slice of a 3D image using an exponential function (for example, using a Tofts model). The parameters that define the exponential function can then be averaged across diseased regions and longitudinal variation in the median values studied.


The oomph-lib logo

Previous to this (2007-2009), I worked on the oomph-lib project in the School of Mathematics at the University of Manchester, in particular on parallel domain decomposition techniques and associated communication requirements.

I completed my PhD at the School of Earth Sciences, University of Leeds, in May 2007.


Recent publications

For a complete list, see my ORCID profile.

Recent (first-author) conference presentations


Past publications