FDA clears GE's AI-powered x-ray algorithms for triaging collapsed lungs

Dive Brief:

GE Healthcare has received 510(k) clearance for AI algorithms designed to detect a type of collapsed lung on x-ray images, the company said Thursday.
The FDA clearance covers a set of algorithms that GE has embedded in its Optima XR240amx mobile x-ray device, thereby enabling hospitals to access AI capabilities without investing in IT infrastructure.
GE, which claims the system is the first AI-powered x-ray device, said the technology will help hospitals identify high-priority cases and fast track their treatment.

Dive Insight:

Turnaround times for chest x-rays vary significantly, even within healthcare institutions. A retrospective chart review conducted in 2018 showed that out of 1,000 chest x-rays ordered with STAT priority, 38% of studies did not indicate clinical urgency, suggesting the status is overused. That results in the x-rays of people with life-threatening conditions being mixed in with those of people with less pressing health problems.

GE said AI can help improve the prioritization of x-rays. The algorithms packaged in the Critical Care Suite that recently received 510(k) clearance from FDA automatically review x-ray images for signs of pneumothorax, a type of collapsed lung. If the algorithms identify a likely case of pneumothorax, an alert is sent to the radiologist. The technologist also receives an on-device notification.

Another set of quality focused AI algorithms automatically rotate images and assess whether there are problems, such as field of view errors, that will make it harder to diagnoses a patient’s condition.

In the tests that supported the 510(k) filing, GE found the algorithms achieved an overall area under the curve of 0.96, suggesting the test is highly accurate. The algorithms are particularly good at identifying large pneumothoraces but also performed well on smaller cases, achieving an area under the curve of 0.94 in the latter population.

The introduction of the algorithms is unlikely to have a big impact on radiology departments, given that pneumothoraces make up a fraction of their caseloads. However, the algorithms could result in people being treated faster and, more broadly, represent a milestone in the move to AI-augmented care.

GE wants to play a big role in driving that transition. The company identified AI as a growth driver for the healthcare unit when it proposed spinning it off last year. GE later dropped that plan, at least for now, in favor of selling its biopharma unit but the focus on AI in healthcare remains.