Mark Hinton, Presenter: Employee, Image Analysis Ltd

Olga A. Kubassova PhD, MSc, Abstract Co-Author: Founder, Image Analysis Ltd Director, Image Analysis Ltd

Mikael Boesen MD, PhD, Abstract Co-Author: Advisor, Image Analysis Ltd

The challenges of handling large image datasets and real-time overlay calculations have been addressed through a novel architecture. Our validation in real clinical practice has shown that our cloud based architecture gives the same or better performance than a workstation. Further it supports multi-center collaboration and seamless data sharing. There are low costs to deploy the software. Development of new functionality is faster and automatically deployed to all users

To achieve efficiency in analysing medical images many radiology units use cloud based computer aided detection (CAD). The problem is to keep calculations and image overlays up to date whilst providing good user experience across bandwidths and latencies that are not controlled. Further, to support new developments, the architecture of the software must support easy integration of algorithms without compromising performance. We present a novel approach to multi-tier architecture, Dynamika, which has successfully addressed the problems and been validated in radiology practices.

The architecture of Dynamika makes use of a classic back end framework of Spring and Hibernate to give robust server side scaling and performance. It uses Spring Webflow to control the path through the application. Webflow has been enhanced to allow for tightly controlled batch processing, which is utilized in clinical trials or routine analysis. The front end is based on Google Web Toolkit to give high performance in the client, desktop like behavior through AJAX and the power of HTML5. 3D visualization and animation is achieved through WebGL.

Software using the new architecture has been bench marked against a conventional workstation solution for user experience and development efficiency. The performance of the cloud  is comparable or better than the workstation in scrolling  images with complex overlays and making calculations such as image registration, saving clinician time. To implement new algorithms, which was measured by recording time of code and test, was up to 10 times less in the cloud architecture. The cloud architecture properly supports collaboration and sharing and supports any device with network access.

Hinton, M, Kubassova, O, Boesen, M, An Enterprise Class Computer Aided Detection Platform Scalable from Laptop to Cloud.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13016507.html