Innovation in image analytics

Medical
imaging has evolved over the years and is now used to provide interventional
treatments, such as stent insertion for heart attacks and to screen for early onset cancers.
Medical imaging can provide information about tissues adjacent to areas of abnormality,
helping clinicians decide whether certain treatments are likely to be of benefit or may cause
unwanted side effects. As imaging techniques have developed to provide greater detail, the
size of the images files has also increased significantly. This has an impact on storage costs
and computer performance required to process these images. Automated computer analysis of
medical images can highlight abnormal areas but also identify those areas that are likely to be
of clinical relevance and need further attention. These techniques have been used to support
radiologists’ workload and have been shown to improve accuracy of diagnosis.
There is a large amount of research being undertaken to determine new ways of analysing
images to enable early detection of disease and to develop new biomarkers and drugs that will
provide treatments targeted only at areas of abnormality. Achievement of these objectives
requires the ability to have a platform to analyse clinical and research images at speed and in
detail, without impacting on the performance of the IT infrastructure already in place.
Our proof of concept proposal aims to take each image and separate out the descriptive
information from the image data so these can be stored separately. Images will be broken up
into constituent pixels and distributed on a newer type of computing infrastructure, which can
store huge amounts of data. This computing platform will also provide tools to enable
researchers to model pixel data so that they perform complex analyses. We feel this platform
will provide a novel way of developing new models for analysing image data that can be
implemented back into the clinical environment to enhance current diagnostic techniques.