CT Slice: Deeper Analysis — Radiomics Study Probes Cardiovascular Disease Risk in Underserved Population
By Keith Loria
Vol. 22. No. 3 P. 26
A new study has found that radiomics—the extraction of highly detailed quantitative features from medical images—can provide a more precise understanding of how cocaine use and other risk factors affect the course of atherosclerotic cardiovascular disease (ASCVD). Shenghan Lai, MD, MPHR, lead author of the study, notes that the research reveals the power of radiomics to improve the understanding of cardiovascular disease, cancer, and other conditions as well.
Since 1999, Lai, a professor at Johns Hopkins Bloomberg School of Public Health, has received continuous funding from the National Institute on Drug Abuse to investigate the effects of HIV, antiretroviral therapy, cocaine use, and other traditional risk factors on subclinical coronary atherosclerosis among African Americans.
“Before 2004, the main outcome was coronary calcification since contrast-enhanced coronary CT angiography [CCTA] was not available before then,” Lai says. “We found long-term cocaine use and antiretroviral therapy were associated with significant (≥50%) coronary stenosis.”
ASCVD typically develops over time as plaque builds up inside arteries. This process, known as atherosclerosis, can eventually lead to life-threatening events such as heart attack and stroke.
Imaging techniques such as CCTA traditionally have offered information on atherosclerosis by describing the degree of stenosis, or narrowing, in the coronary arteries. However, while these measures are useful, physicians report that they aren’t always the most precise way to assess the risk of an adverse event such as a heart attack.
From 2004 through 2015, Lai and his research team collected CCTA data from approximately 1,429 study participants. This led to a paper published in the Journal of the American Heart Association, “HIV Infection Itself May Not Be Associated With Subclinical Coronary Disease Among African Americans Without Cardiovascular Symptoms.”
“Our study demonstrated that HIV alone was not associated with the presence of subclinical coronary atherosclerosis,” he says. “Inspired by a paper published in Radiology, I started to explore how to use radiomics to study whether and how HIV, cocaine, and conventional risk factors influenced the progression of subclinical coronary atherosclerosis.”
In early 2018, Lai read a paper by Márton Kolossváry, MD, a research fellow at the Cardiovascular Imaging Research Group of Semmelweis University in Budapest, Hungary, who is a pioneer in performing radiomic analysis in coronary atherosclerosis, and visited his institution.
“I met his mentor and the director of his institute, presented our data, and hoped to develop collaborations,” he says. “I hired Dr. Kolossváry as a post-doc fellow at Johns Hopkins in late 2018, and he stayed in Baltimore until early 2020. We have continuously worked to develop more papers. The purpose of bringing Dr. Kolossváry to our program was to apply radiomics to investigate the contributions of HIV, cocaine, and other factors to subclinical coronary atherosclerosis.”
In the Johns Hopkins study, each of 1,276 radiomic features was treated as the outcome variable in a linear mixed model, and potential confounding factors were adjusted for. Lai explains that from a radiomic perspective, one of the chief findings of this study was that cocaine use may contribute to disease progression more than HIV among the African American population.
“According to multivariate analysis, while 32% (409/1276) of all radiomic parameters showed significant association with cocaine use, HIV infection, and elevated ASCVD risk, cocaine use and HIV infection were significantly associated with 23.7% (303/1276) and 1.3% (17/1276) of the radiomic features, respectively, suggesting cocaine may contribute more to the disease progression than HIV,” he says. Additionally, the study found that HIV, cocaine use, and elevated ASCVD risk may make certain independent contributions to disease progression, as there was no overlap among radiomic features associated with elevated ASCVD risk, cocaine use, or HIV infection.
Before radiomics, anatomical plaque burden severity was the focus in patient care and research, and the characteristics of coronary plaque were not fully appreciated. Lai notes that radiomics provides quantitative information from medical images, especially the nature of coronary plaques, offering incremental information of value to the identification, classification, and prognosis of disease.
“In our case, radiomics examined the characteristics of coronary plaques and provided a more complete understanding of how cocaine use and other risk factors affect the course of coronary artery disease,” he says.
For ASCVD, especially subclinical coronary artery disease, Lai says contrast-enhanced CCTA is essential to the diagnosis and monitoring of disease development and progression.
“CCTA provides critical information for disease diagnosis and management. Radiomics utilizes images generated from CCTA and provides additional information,” Lai says. “Physicians who use medical images for diagnosis and prognosis should be informed that radiomics may provide important additional information about the disease diagnosis and management. Without radiomics, the characteristics of coronary plaques may not be fully appreciated and utilized. Especially, some characteristics hidden in high-dimensional interactions in the images may be overlooked. With Big Data analytics approaches, the hidden information may provide critical prognostic data of value to physicians and patients.”
The Future of Radiomics
Lai’s study was a pilot investigation, and he is planning to perform further radiomic analyses to confirm the findings.
“There are two compelling reasons for us to continue this study,” he says. “The question of whether HIV is associated with coronary artery disease is not only a scientific issue but also a societal one. We are interested in learning whether HIV is a causal factor of coronary artery disease. We are also interested in learning the contributions of HIV and cocaine use to the progression of coronary atherosclerosis among the underserved African American populations. The contribution of drug use, especially cocaine, in the pathogenesis of cardiovascular disease is grossly underappreciated, especially in these populations.”
Radiomics has two components: extracting data and performing Big Data analytics with machine learning. This process will likely play an important role in the future of health care.
“The data extracted from medical images, combined with genome, clinical, and other data could be combined for developing precision medicine with the use of an artificial intelligence approach—some people call it radiogenomics,” Lai says. “Whenever a physician requests radiological evaluation, he or she should consider whether radiomics could be helpful.”
— Keith Loria is a freelance writer based in Oakton, Virginia. He is a frequent contributor to Radiology Today.