AI-embedded X-Ray system could help speed up detection of a collapsed lung

Pneumothorax impacts nearly 74,000 Americans every year. An AI algorithm could help radiologists prioritize the review of critical cases


With more than 2 billion X-Ray exams done annually, X-Ray is often the hospital’s first impression of a patient. Just like first impressions with people, the first image taken helps set the path going forward. “We are getting portable X-Rays all the time for our patients,” said Dr. Rachael Callcut, Associate Professor of Surgery at the University of California, San Francisco (UCSF) Medical Center and Director of Data Science for the Center for Digital Health Innovation. “When an X-Ray is taken on a patient, especially a patient who's suffering from an emergent condition or a potentially life-threatening condition, the time that it takes to process, have someone read that and have the image actually come into a queue is a really important time period where minutes and hours matter.” For example, a collapsed lung, known as a pneumothorax, is a condition which strikes nearly 74,000 Americans each year [1] and can be deadly if not diagnosed quickly and accurately [2]. A pneumothorax occurs when air leaks into the space between the lung and chest wall. This air pushes on the outside of the lung and makes it collapse. It can be caused by trauma, cigarette smoking, drug abuse, certain lung diseases or be caused by complications from surgery.

Today, patients who present with symptoms associated with this condition receive a chest X-Ray, which can take anywhere between two to eight hours to read[3]. Tension pneumothorax or an enlarging pneumothorax can develop as a result of delayed treatment [4], potentially leading to fatal consequences if not treated quickly. This is an example of what may be designated as a “STAT” chest X-Ray, which is supposed to be reserved for potentially life-threatening circumstances. It is a designation on the exam placed at the time of order entry and refers to the ordering provider’s determination that the results require immediate interpretation and follow-up. STAT portable chest X-Rays can account for more than 60 percent [3] of a radiology center’s mobile chest X-ray volume, almost double that of routine exams. “There's no universally accepted definition of what constitutes a STAT exam,” said Dr. Karl Yaeger, a diagnostic radiologist at St. Luke's University Health Network in Bethlehem, Pennsylvania. “Is it STAT because the patient is medically unstable? Is it STAT because they're pending discharge? Or is it STAT because it's a patient that's demanding an immediate answer? Any of those could constitute a STAT exam. That is difficult for radiologists to know on the other end when they see that exam. We are trying to prioritize our work lists with our technology to make the most urgent findings that may exist on a study rise to the top of a STAT work list to have those interpreted first.” That’s why clinicians are looking for opportunities to read STAT chest X-Rays faster and in a more prioritized manner to enable a quicker diagnosis. One such opportunity is the Critical Care Suite* on Optima XR240amx, which is designed to identify cases with the critical condition of pneumothorax at point-of-care to enable prioritization of image review. [video width="1920" height="1080" mp4=""][/video]   “The concept behind this was to develop an algorithm using artificial intelligence (AI) that could learn how to find pneumothorax on a chest X-Ray and draw attention and alert a clinician and radiologist to a case that presents a potential finding,” said Dr. Callcut. “And by doing so, it could allow us to actually speed up the timely diagnosis of a potentially life-threatening condition.” Critical Care Suite will employ a suite of AI algorithms, such as pneumothorax detection, designed to identify this potentially life-threatening condition in chest X-Rays with high accuracy [5]. Additionally, quality-focused AI algorithms simultaneously analyze and flag protocol and field of view errors as well as auto rotate the images on device for technologists. The AI algorithms are hosted on the mobile X-Ray system – a first of its kind AI-embedded imaging device – designed to share the output through an onscreen notification. Critical Care Suite and the quality algorithms on Optima XR240amx are powered by Edison a next generation intelligence platform that helps accelerate the development and adoption of AI technology and empowers providers to deliver faster, more precise care. “I think the idea and concept about an on-device alert really gets to the heart of early warning detection technology,” said Dr. Callcut. “There are many opportunities to use early alerts and early warnings and it's very clear that the sooner the clinician knows of a potential life-threatening or major finding, the more likely they are to be able to do a timely intervention that could perhaps change the trajectory of a patient.” When a pneumothorax condition is identified, an alert, along with the original chest X-ray - is sent directly to the radiologist for review via picture archiving and communication systems (PACS). The technologist also receives a subsequent on-device notification [6] to give awareness of the prioritized cases. “The benefit of having AI algorithms is ultimately providing better patient care,” Dr. Yaeger said. “If we can decrease the amount of time it takes to detect a critical finding, communicate that finding appropriately and lead to more expedient treatment, the patient benefits and ultimately, we as an institution benefit.” The Critical Care Suite was developed with the goal of improving timely patient care through a collaboration between GE Healthcare and the UCSF Center for Digital Health Innovation. Dr. Callcut partnered with Dr. John Mongan and Dr. Andrew Taylor, radiologists at UCSF, to create the initial use case and data science approach behind the pneumothorax detection algorithm. For the algorithm validation study, UCSF and St. Luke’s University Health Network – along with Humber River Hospital in Toronto, Canada, and Mahajan Imaging in New Delhi, India – worked alongside GE Healthcare to replicate the initial work carried out in acquiring and annotating images. [video width="1920" height="1080" mp4=""][/video]   “It is our hope that the AI platform will expedite the review of studies so that the radiologist may appropriately communicate emergent findings to the appropriate doctor in a minimal amount of time,” Dr. Yaeger said. “The patients are ultimately benefited by more expedient and more accurate diagnosis of life-threatening conditions.” Dr. Yaeger hopes solutions like the Critical Care Suite will ultimately improve radiology by enabling radiologists to do their jobs better with more accuracy and efficiency. “Going forward, I hope that there comes a day where critical findings such as large pneumothorax are always detected within minutes,” Dr. Yaeger said.     [1] [2] [3] Rachh, Pratik, et al. “Reducing STAT Portable Chest Radiograph Turnaround Times: A Pilot Study.” Current problems in diagnostic radiology (2017). [4] Lorenz, Jonathan, and Matthew Blum. “Complications of percutaneous chest biopsy.” Seminars in interventional radiology. Vol. 23. No. 2. Thieme Medical Publishers, 2006 [5] Critical Care Suite’s overall Area Under the Curve (AUC) for detecting a pneumothorax is 0.96. Large PTXs are detected with extremely high accuracy (AUC = 0.99). Small PTXs are detected with high accuracy (AUC = 0.94). GE Healthcare 510k K183182. [6] The technologist on-device notification is generated after a delay, post exam closure, and it does not provide any diagnostic information, nor is it intended to inform any clinical decision, prioritization, or action.