GE Healthcare, Boston Children’s join to aid pediatric scans
Boston Children’s Hospital is working with GE Healthcare to develop decision support software to help physicians interpret brain scans of children more quickly and accurately.
Children pose a particular challenge to radiologists, as rapid changes in the body occur during early development, making it more difficult to differentiate between normal and abnormal scans. What is a normal brain scan for a child’s age soon changes, resulting in large variability of brain MRI scans, says Richard Robertson, MD, radiologist-in-chief at Boston Children’s.
The joint project involves building and validating multiple algorithms that learn all the variables and can give physicians a set of normal brain patterns and a set of abnormal patterns to use as a benchmark when reading scans, so physicians know where they want to start with the scanning process.
“Since most pediatric imaging is not performed in children’s hospitals by specialists, this new digital tool, once available, will provide non-specialists with access to knowledge and expertise to help effectively diagnose children,” Robertson explains. “We believe that by providing decision support at the time of interpretation, we can improve both the confidence and performance of the interpreting radiologist.”
When conducting a study, there can be five or six ways of looking at the brain, according to Robertson. Having access to sets of normal and abnormal patterns lets a doctor decide if a particular image shows what would be expected, and, if it is not what is expected, “what type of abnormality we are talking about,” he adds.
The idea for creating algorithms to support medical imaging interpretations came about 18 months ago, Robertson notes, but the project just got going about six months ago, after resources were found. The first phase is to develop a normative reference database to improve diagnostic accuracy and the speed of reading exams.
GE Healthcare is co-creating the algorithms and is providing a cloud-based platform to host the images and generate the computing power that physicians will use to compare scans when interpreting a study.