Microsoft Research: Medical Imaging Search Engine

Right now, we're living in a world where we look at more computer generated imagery than anything else. Doctors are reaching a critical point where the amount of medical imagery generated during something like a routine CT scan is daunting to navigate. Kenju Suzuki at the University of Chicago says, "As medical imaging has advanced, so many images are produced that there is a kind of information overload. The workload has grown a lot." Antonio Criminisi leads a group at Microsoft Research in Cambridge, U.K. working on a system that will make it easier for doctors to work with databases of medical imagery. The system indexes the images generated during the scans. It automatically recognizes organs, and they are working to train the system to detect certain kinds of brain tumors.

Criminisi's team has developed a piece of software that takes the collection of 2-D and 3-D images and indexes them together. There is a short demonstration video on MIT's post where he describes a number of the different features currently available in this piece of software. Currently it can give doctors the ability to quickly search and index a database of images by organ. The system puts the indexes together through an analysis of the images taken. Currently, medical imaging databases use the text comments linked to the image for doctors to search. This gives them the ability to search, but it takes time because not all of the results are relevant. These kinds of systems will allow doctors to easily navigate from new images to old images in the same patient, side-by-side. It will also allow doctors to easily pull up images from other patients for comparison.

The team is also working on integrating the technology found in Microsoft's Kinect. This will give surgeons the ability to navigate through the images with gestures. This will give them access to the images mid-procedure without them having to touch a mouse, keyboard, or even a touch screen. As these are all things that could compromise the sterility of the operation, this will be a very useful tool. Criminisi's team plans for this tool to be implemented at a large scale, making automatic indexes of images as they are scanned and tying them into the greater database seamlessly.

[via MIT: Technology Review]