Computer Vision Startup Oxipit Automates Chest X-Rays
By Adam Pease
This week, a regulatory committee for the European Union greenlit the use of a new radiology tool called ChestLink. The tool, developed by Oxipit, uses computer vision to automate the process of reviewing chest x-rays. This blog discusses the news and what it could mean for the future of both computer vision and healthcare.
Who Is Oxipit? Healthcare Gets Automated
Oxipit’s ChestLink has just been granted a CE mark in the EU, which clears it for the safety of the device for mainstream use, but it does not expect the tool will see use until 2023. Still, it has already been operating in supervised settings where it has been processing nearly half a million chest X-rays over the past year. According to its own metrics, ChestLink has made zero errors that are clinically relevant in its trials, but the next step will be to bring it to actual markets.
While it has performed well in supervised settings, Oxipit aims to deploy ChestLink without radiologists present. Rather than having humans analyze X-ray images, Oxipit’s tool will first scan the images, checking for signs of abnormality, and then pass on those scans to a radiologist if any issues are flagged. Because most X-rays scanned by human beings have no issues, this approach to automation has the potential to eliminate a considerable amount of work in the medical imaging pipeline. Still, it remains to be seen whether patients and doctors will appreciate the technology or not.
Computer Vision Can Revolutionize Care
While the announcement has been met with consternation by career radiologists, it could create a revolution in care for patients—both in terms of cost and accessibility. For many, the challenge of seeing specialists comes from difficulty finding an appointment in their busy schedule or from the high cost of services. In some parts of the EU, healthcare access is gatekept by primary care referrals, but the expansion of access to radiology could mean that these barriers are removed and that the general cost of radiology services will decrease with less scarce treatment options.
At the same time, stories like the downfall of Theranos have made doctors, patients, and investors cautious about the promises and risks of self-service healthcare. The risks of an AI-enabled system that misses a critical abnormality in an X-Ray cannot be understated. Still, examples like this take advantage of the strengths of computer vision and seem to perform very well in test settings.
For many patients, there will be no comparison to meeting with a real doctor. However, the convenience and potential affordability of automated solutions like Oxipit’s may sway the public to depend more and more on AI for healthcare services. The rollout of ChestLink in Europe will be a test case for the adoption of computer vision solutions and will reveal much about the public’s willingness to put their health in the hands of AI.