Ashish Sharma, Presenter: Nothing to Disclose

Yusuf Nadir Saghar MS, Abstract Co-Author: Nothing to Disclose

Imaging research has usage patterns that change and are very different from clinical usage patterns. NOSQL and clouds can provide researcher the flexibility to define what to search and alleviate the burden of complex infrastructure.

Researchers often store DICOM images in a PACS or in relational databases. For performance, they use a fixed schema, which in turn limits the number of DICOM headers that can be queried. This limits the type of research queries that can be executed on such systems. NOSQL are a relatively new class of document-centric schema-free databases. These systems store metadata as key-value pairs with the actual data stored as binary blobs on the file system. This design is well suited to research image management, providing users the flexibility to define and modify the attributes that they query Another challenge faced by researchers is the need for complex IT infrastructure to manage ever-growing size and complexity of collected data. This problem is solved by Cloud platforms like Amazon. Here we describe our efforts to utilize NOSQL storage and cloud platforms for medical image management 

NOSQL databases offer the flexibility to index and query all metadata. This increases the types of queries a user can run and makes them ideally suited for research in areas such as dose, & quantitative imaging. Clouds allow users to reduce their IT burden and focus on science

We use MongoDB as the image database. Files are parsed and metadata is stored as key-value pair. Image series are stored on file system with links in MongoDB. An Apache CXF based webservice allows external users to query/retrieve/submit images via https. We have modified our VirtualPACS so remote clients can also use existing DICOM workstation to interact with our NOSQL archive using DICOM messages. The webservice is responsible for authorizing and auditing user access. The system is deployed on Amazon EC2 and it allows users to scale on-demand.Performance tests were run by uploading DICOM collections from TCIA. Clients then queried and retrieved images from various Amazon EC2 locations. The system was compared against a DICOM archive with clients accessing it over VPN

Sharma, A, Saghar, Y, A NOSQL Based Research: PACS Deployed on a Cloud.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12031184.html