Single chest X-ray can predict ASCVD risk, determine statin eligibility
Click Here to Manage Email Alerts
A deep-learning model using information from a single chest X-ray reliably predicted atherosclerotic CVD risk and statin eligibility when compared with a common risk calculator, researchers reported.
The findings were presented at the Radiological Society of North America annual meeting.
“Clinicians usually use a chest X-ray to make a diagnosis, like pneumonia or a lung mass,” Michael T. Lu, MD, MPH, co-director of the Massachusetts General Hospital (MGH) Cardiovascular Imaging Research Center, associate chair of imaging science at the MGH department of radiology and assistant professor of radiology at Harvard Medical School, told Healio. “The whole idea with this research is that there is additional information on the images about someone’s future risk for cancer or heart disease that we do not see as radiologists, or perhaps see but don’t know how to categorize it. The idea was to use existing trial data, feed it into an [artificial intelligence] algorithm along with what happens to a person 6 to 12 years later, and teach that tool to predict which patients go on to have a heart attack or stroke.”
Using AI to predict CV risk
Lu, Jakob Weiss, MD, a radiologist with the Cardiovascular Imaging Research Center at MGH and the AI in Medicine program at the Brigham and Women's Hospital, and colleagues developed a chest X-ray (CXR)-CVD risk model using 147,497 CXRs from 40,643 participants from the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer screening trial. The model was trained to predict CV mortality from a single CXR image. Independent testing was performed in a second separate cohort of 11,430 outpatients potentially eligible for primary prevention who had an LDL between 70 mg/dL and 190 mg/dL, no diabetes and no prior major adverse CV events (mean age, 60 years; 42.9% men).
Researchers defined statin eligibility as a 10-year major adverse CV event risk of 7.5%.
Researchers then compared the prognostic value of the CXR-CVD risk model with the established ASCVD risk score in the subset of 2,401 patients who had the necessary variables to calculate ASCVD risk. The primary outcome was observed 10-year incident major adverse CV events, defined as MI and stroke.
“There are currently established guidelines from the American Heart Association and the American College of Cardiology to establish statin candidacy for primary prevention of CVD,” Weiss told Healio. “Currently, this risk estimation is based on a relatively complex risk calculator that takes nine different variables as input and, oftentimes, these variables are not available for the patient. Why not look whether this information is somehow embedded in a chest X-ray and see if we can use this as the only input to estimate risk for future CV events?”
During median follow-up of 10.3 years, there were 1,096 major adverse CV events observed in the independent testing dataset.
‘Opportunistic risk assessment’
There was a significant association of CXR-CVD risk and major adverse CV events among statin-eligible patients (HR = 2.03; 95% CI, 1.81-2.3; P < .001), which persisted after adjustment for CV risk factors (adjusted HR = 1.63; 95% CI, 1.43-1.86; P < .001).
Among patients with the necessary variables to calculate ASCVD risk, the performance of the CXR-CVD risk model was similar to the ASCVD risk score (C statistic, 0.64 vs. 0.65; P = .48), which persisted after adjustment (aHR = 1.58; 95% CI, 1.2-2.09; P = .001).
Both Weiss and Lu said the CXRs could serve as “opportunistic risk assessment,” where patients who have higher ASCVD risk based on the X-ray could potentially be referred to a cardiologist for further evaluation.
“We are not recommending people get chest X-rays,” Lu said. “Chest X-ray is one of the most common medical tests. If someone happens to get one for another reason, say a fever or cough, or preoperative planning for another procedure, this tool could run in the background of the electronic health record and perhaps identify people who might be eligible for a statin. That is the idea of this project.”
Reference:
- AI predicts heart disease using single X-ray. Available at: press.rsna.org/timssnet/media/pressreleases/14_pr_target.cfm?id=2388. Published Nov. 29, 2022. Accessed Dec. 5, 2022.
Collapse
Weiss J, et al. Abstract #TS-SSCH04-1. Presented at: Radiological Society of North America Annual Meeting; Nov. 27-Dec. 1, 2022; Chicago (hybrid meeting).
Click Here to Manage Email Alerts