Opinion: New FDA-cleared X-ray AI delivers when timing is critical

Minutes matter when dealing with a collapsed lung

By Tom McGuinness

Minutes matter when dealing with a collapsed lung. Every minute spent scanning the patient matters; every moment processing the medical image matters; every hour that passes before an image is reviewed matters – and it’s vital the process is as fast as possible. For years the industry has promised that artificial intelligence (AI) would eliminate delays and accelerate medical processes like the diagnosis of collapsed lung. Finally, we’re beginning to see that promise of AI spring to life... in the same area medical imaging was born more than a century ago: the X-ray. Pneumothorax, or collapsed lung, is a life-threatening condition with potentially devastating effects – unless it’s caught early. It affects 74,000 Americans and hundreds of thousands of patients worldwide each year. Caused by trauma, smoking, drug abuse or lung disease, a pneumothorax occurs when air leaks between the lung and chest wall, pushing against the lung and making it collapse. And in almost all cases, collapsed lung is diagnosed first by an X-ray. Alongside clinical research partners like UC-San Francisco Medical Center, we’ve developed an AI solution called Critical Care Suite to help radiologists prioritize cases and to provide tools for technologist to improve exam quality. These next-gen AI algorithms automatically and almost instantaneously scan chest X-rays and flag pneumothorax with impressive accuracy. The AI then triggers immediate alerts to the radiologists to escalate the review high-priority images which could help expedite treatment. This could drastically cut the average review time, which is currently up to eight hours.[1] It’s the first step in a long journey toward AI-integrated radiology and enhanced care, and it’s noteworthy as this is the first-of-its-kind AI-embedded X-ray device for triage that has received U.S. FDA clearance. Yet as proud as we are of this milestone, these X-ray algorithms are just the beginning. We see a near-term future for health care and radiology where AI, deep-learning and digital analytics augment every medical imaging task across the care continuum and work seamlessly across devices and applications, and where clinicians are empowered to deliver better, more personalized care that improves patient outcomes within an ecosystem of precision health. This future is why we continue to invest in our Edison intelligence portfolio. What we’ve done with X-ray we’re doing across the medical imaging portfolio as well, because doctors and patients rely on a variety of diagnostic tools to treat diseases. GE Healthcare alone has more than 500,000 intelligent scanners ready for AI integration, and we’re adding AI to those on a variety of fronts. In MR imaging, our new AIRx workflow tool uses deep-learning algorithms to automatically “prescribe” slices and help clinicians reduce redundant, manual steps. In prenatal ultrasound, our unique SonoCNS measures fetal brains by aligning the system automatically and reducing sonographers’ keystrokes by more than 75 percent. For CT scans, our deep learning reconstruction algorithm produces true fidelity images. We’re delivering on the future promise of health care AI today, by augmenting clinical decisions and improving patient outcomes in the minutes and moments that matter. Where do you see potential for AI to change health care for the better?     [1] Rachh, Pratik, et al. "Reducing STAT Portable Chest Radiograph Turnaround Times: A Pilot Study." Current problems in diagnostic radiology (2017).   Tom McGuiness is President and CEO of GE Healthcare, Imaging. This post originally appeared here. Follow Tom @tom_mcguinness