Medical Image Querying Meeting

Arising from meeting on 15/3/06 with:

• TimCootes^?
• ChrisRose^?
• ChrisTaylor^?
• DanieleTuri
• AlanWilliams
• TomosWilliams^?
• KatyWolstencroft
• GeoffParker^?
• CharlesHutchinson^?

Minutes

Things that people thought would be useful/interesting:

• To be able to show the 'pipeline' of processes that have been applied to a particular image, i.e. to get a log of what has happened to it. ChrisTaylor^? suggested this should be a web page that allows one to 'drill down' into greater details.
• Be able to re-create the entire analysis, right down to software versions.
• Be able to log all user interaction, including all button clicks and their order.
• To be able to answer patient queries under the data protection act, e.g. "what did you do to my data?".
• To be able to know about data duplication (who, when etc.).
• To be able to relate the results of image analysis experiments to patient outcomes (e.g. hospital admissions).
• To be able to interface to the ElectronicPatientRecord^?.
• To be able to automatically generate anual progress reports for ethics committees, e.g. "how many participants so far, what has happened to them? etc." (See www.corec.org)
• To be able to compare methodologies. E.g. run someone else's algorithm in place of your and compare results. However, there are issues with this: they may not be allowed to see the data you are sending them for analysis. So, how do we enforce such contracts (currently we simulate this by using a trusted community).

General points:

• Principal Investigators are responsible for the data their project collects; how does this translate to the web services/grid model?
• Issue of manually-entered metadata (e.g. patient name being entered incorrectly), and moving to a situation where the NHS number of the patient is used and everything is linked to their central records.
• ChrisTaylor^? specified that we currently assume that the ethical approval aspect of security policies is automatically applied.
• Scalability?