The number of CT scans hospitals perform is on the rise, but professionals who actually read these scans can’t seem to keep up with increasing demand — often just looking at images as they come in.
But artificial intelligence may be able to help. Chicago startup Realize.ai is launching a pilot study of its AI that identifies certain problems often spotted on CT scans this fall at Northwestern University Medical Center. Many times, these scans can reveal deadly conditions that need immediate attention, which means reading scans efficiently can be a life-or-death situation.
“Getting to a scan too late or misreading a scan can cause serious clinical problems and we are trying to alleviate that,” said co-founder Alex Risman. “The way we are doing this is trying to apply some new technology called deep learning towards automatically identifying issues in common types of medical imaging like CT scans, X-rays and MRIs.”
Realize hopes to speed the process in three ways. First, it will try to identify the most pressing scans so radiologists can look at them first. The AI will also work to highlight problematic areas of a scan to speed up the process.
In the future, the deep learning algorithm may even act as a second opinion, comparing one scan against a huge library of other scans to pinpoint anomalies and helping radiologists who may have missed something.
Initially, the company is focused on spotting intracerebral hemorrhages (right) — which represent 10 percent of all strokes — and reading chest scans. But in the future, Realize looks to move beyond the medical industry altogether.
“Our broader vision is to create labor-saving technology across many industries and view work in different industries as a template for work in new industries,” Risman said. “One of the beautiful things about this technology, deep learning, is that the same model that might work for identifying a clinically relevant anomaly or pathologic finding in CT scan might also work for identifying a pedestrian in a video feed [of a self-driving car].”
Risman, a data scientist, has been working on this brand of AI since February, but it only took him two months to develop a model that spotted 90 percent of intracerebral hemorrhages. Along with two other co-founders, including a trained radiation oncologist, Realize is gearing up to launch in Matter, the Chicago healthcare innovation incubator, followed by the Northwestern pilot this fall.
Images via Shutterstock, Lucien Monfils/Creative Commons