Apr 8 2020
COVID-19 is a pandemic caused by a new coronavirus infection. For a majority of the patients who died from this disease, pneumonia was the ultimate cause of death.
Chest X-rays from a patient with COVID-19 pneumonia, original x-ray (left) and AI-for-pneumonia result (right). Patient has a pacemaker device and an enlarged heart, which indicates that the AI algorithm is powerful enough to work even when the patient has underlying health issues. Image Credit: UC San Diego Health.
Pneumonia is a medical condition that is characterized by inflammation and accumulation of fluids in the lungs, which causes difficulty in breathing. Acute pneumonia usually needs assistance with ventilators and involves prolonged hospital stays in intensive care units. Ventilators are medical devices that are currently in huge demand in certain cities struggling with the increase of COVID-19 cases.
To rapidly identify pneumonia - and better differentiate between COVID-19 patients who may probably require more supportive care in hospital settings and those who could be closely monitored in home environments - radiologists and other physicians from UC San Diego Health are currently utilizing artificial intelligence (AI) to expand the lung imaging analysis in a clinical research study. This study was facilitated by Amazon Web Services (AWS).
With this novel AI capability, physicians at UC San Diego Health have gained exclusive insights into more than 2,000 images, to date. In one similar case, a patient admitted in the Emergency Department who did not show any COVID-19 symptoms was made to undergo a chest X-ray procedure for other reasons.
However, the AI readout of the X-ray suggested signs of early pneumonia, which was subsequently confirmed by a radiologist. Consequently, the patient was also tested for the COVID-19 disease and was found to be positive.
We would not have had reason to treat that patient as a suspected COVID-19 case or test for it, if it weren’t for the AI. While still investigational, the system is already affecting clinical management of patients.
Christopher Longhurst, MD, Chief Information Officer and Associate Chief Medical Officer, UC San Diego Health
The novel AI capability garnered attention many months ago when Albert Hsiao, MD, PhD, a radiologist at UC San Diego Health and an associate professor of radiology at the University of California San Diego School of Medicine, and his research team created a machine learning algorithm that enables radiologists to apply the AI algorithm to improve their own abilities to detect pneumonia on chest X-rays.
The AI algorithm, trained with 22,000 notations by human radiologists, superimposes X-rays with color-coded maps that denote the probability of pneumonia.
“Pneumonia can be subtle, especially if it’s not your average bacterial pneumonia, and if we could identify those patients early, before you can even detect it with a stethoscope, we might be better positioned to treat those at highest risk for severe disease and death,” stated Hsiao.
In the recent past, Hsiao’s research team employed the AI method on 10 chest X-rays, reported in medical journals, from five patients who were treated in the United States and China for the COVID-19 disease.
The AI algorithm reliably localized the pneumonia spots, in spite of the fact that the images were taken at many different hospitals and differed significantly in terms of contrast, technique, and resolution. The details of the study were published in the Journal of Thoracic Imaging.
Thanks to donated service credits extended by the AWS Diagnostic Development Initiative and the efforts of the Clinical Research IT team at UC San Diego Health, Hsiao’s AI technology has been employed across UC San Diego Health in a clinical research study that enables all radiologists and physicians to get a preliminary estimate about a patient’s chance of having pneumonia in just a few minutes, at point-of-care.
AWS has partnered with us on multiple projects in the past. Once COVID-19 became a crisis, AWS reached out to us and asked if there was anything they could do to help. My mind immediately went to a presentation I’d seen Albert give on their initial AI tests for pneumonia. AWS helped our Clinical Research IT team get the study up and running system-wide in just 10 days.
Michael Hogarth, MD, Professor of Biomedical Informatics, UC San Diego School of Medicine
Hogarth is also the clinical research information officer at UC San Diego Health. Hsiao added that the chest X-rays are more cost-effective, the equipment is easier to clean and more portable, and the results are obtained more rapidly when compared to that of many other diagnostics.
For example, polymerase chain reaction-based clinical diagnostic tests that are used to detect the virus responsible for the COVID-19 disease can take a number of days in certain areas of the United States.
“That’s where imaging can play an important role. We can quickly triage patients to the appropriate level of care, even before a COVID-19 diagnosis is officially confirmed,” added Hsiao.
To make things clear, experts at UC San Diego Health emphasized that they are not simply diagnosing the COVID-19 itself through lung imaging. Many different types of viruses and bacteria can cause pneumonia. Moreover, the use of Hsiao’s AI algorithm is still regarded as an investigational tool. While clinicians can use these images, patient care continues to be guided by formal understanding from human radiologists.
As we prepare for a potential surge in patients with COVID-19, it’s not just patient rooms and supplies that may become limited, but also physician and staff capacity. So it’s tremendously helpful to have tools that allow physicians who are not as experienced as radiologists in reading X-rays to get a quick idea of what they’re looking at, especially frontline emergency and hospital-based physicians.
Christopher Longhurst, MD, Chief Information Officer and Associate Chief Medical Officer, UC San Diego Health
The UC San Diego Health team is now hoping to expand the AI-powered analysis for identifying pneumonia at four other academic medical centers at the University of California.
“As an academic medical center, we’re always looking for ways to bring innovations to the bedside,” Longhurst added. “Although we need more studies to evaluate the effectiveness of this algorithm and improve its accuracy as we see more patients, what we’re seeing so far is evidence that this approach could be a powerful tool for health care providers to provide more reliable, early diagnoses of COVID-19 and other infections.”
Brian Hurt, MD, and Seth Kligerman, MD, from the Department of Radiology of UC San Diego School of Medicine, co-authored Hsiao’s Journal of Thoracic Imaging study.
The study was partly supported by the National Institutes of Health (T32 Institutional National Research Service Award), American Roentgen Ray Society, and NVIDIA Corporation (GPU grant).