In a pair of upcoming conference papers, MIT researchers describe a machine-learning algorithm that can register brain scans and other 3-D images more than 1,000 times more quickly using novel learning techniques. This process usually takes some hours to scan using the traditional system that meticulously aligns each of potentially a million pixels in the combined scans.
Medical image registration is a method that involves overlaying of two images, such as magnetic resonance imaging (MRI) scans that helps to analyze anatomical differences in great detail, for instance, a patient with a brain tumor can be easily assessed. The Doctor will overlay a brain scan that has been done months ago on a scan that has been done recently to analyze small changes in the tumor’s progress.
Thousands of images are paired and registered in the algorithm; it acquires information about how to align images and estimates some optimal alignment parameters. It uses the parameters to map all pixels of one image to another thus reducing time by a minute or two or less than a second using a GPU with comparable accuracy to state-of-the-art systems.
“If you’re able to learn something from previous image registration, you can do a new task much faster and with the same accuracy.”- says co-author on both papers Guha Balakrishnan, a graduate student in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Department of Engineering and Computer Science (EECS).
At the conference this week the paper will be presented on Computer Vision and Pattern Recognition (CVPR), and in September it will be held at the Medical Image Computing and Computer Assisted Interventions Conference (MICCAI).
The MIT Machine Learning algorithm is basically on numerous 2D images that are stacked to form massive 3D images called ‘volumes’ which may contain more than a million 3D pixels known as ‘voxels’. It is sought to be time-consuming to align all voxels in the first volume with those in the second.
You have two different images of two different brains, put them on top of each other, and you start wiggling one until one fits the other. – Dalca, senior author on the CVPR paper and lead author on the MICCAI paper
Dalca finds this procedure time consuming as the optimization takes a long time. This process becomes particularly slow when analyzing scans from large populations.
In the CVPR paper, the researchers trained their algorithm on 7,000 publicly available MRI brain scans and then tested it on 250 additional scans. It was found that algorithm could accurately register all of their 250 test brain scans within two minutes using a traditional central processing unit, and in under one second using a graphics processing unit.
Technology has been long helping human lives and it proves it once more as Dalca says that surgeons could potentially register scans in near real-time, getting a much clearer picture on their progress. Overlapping of an image during a surgery is impossible as it might take long hours, but with technology in handy, it might make it feasible.