The majority of asymptomatic subjects in the MiHEART (Miami Heart Study) had plaque detectable by artificial intelligence–guided quantitative coronary computed tomography angiography (AI-QCT), though the number was far less in human assessment. These were the findings of a new analysis of 2,301 community-based subjects from the ongoing observational prospective cohort study at Baptist Health, Florida, who were apparently free of cardiovascular disease. The study was published Monday online in JACC: Cardiovascular Imaging. Coronary computed tomography angiography (CTA)-derived plaque burden is associated with the risk of cardiovascular events and is expected to be used in clinical practice, said the researchers, led by Keishi Ichikawa, MD, PhD, from Harbor–UCLA Medical Center, CA. They stressed that, “Understanding the normative values of computed tomography-based quantitative plaque volume in the general population is clinically important for determining patient management.” In particular, the researchers said that the introduction of AI–guided quantitative coronary computed tomography angiography (AI-QCT) analysis “has facilitated the accurate and rapid quantitative analysis of coronary plaque volume and characteristics.” The current study therefore set out to investigate the distribution of plaque volume in a general population cohort using the AI based QCT technique. Quantitative assessment of baseline plaque volume was performed using a US Food and Drug Administration (FDA)-cleared AI-QCT for participants with available coronary CTAs and clinical records, and percentiles of the plaque distribution were estimated with nonparametric techniques. Quality assurance of all AI-QCT results was provided by trained radiologic technologists. All participants underwent coronary CTA using a 256-slice volumetric multidetector CT scanner, and the mean number of segments was 14.4 ± 1.0 (median: 14) for participants included in the final analysis. At baseline, the subjects’ mean age was 53.5 years (± 6.7), sex was evenly weighted (50.4% male) and participants tended to be younger (12.7% were 40 to 44 years, 20.3% were 40 to 49, 24.1% were 50 to 54, 22.4% were 55 to 59, 18.2% were 60 to 64 and 2.3% were 65). ‘Disease overcalling’ Overall, 93.1% of men and 84.6% of women had detectable plaque using AI-QCT, and just 11.1% of subjects did not have any plaque (TPV = 0 mm3 ), including 6.9% of male subjects and 15.4% of female subjects. However, CTA analysis performed by human readers returned lower prevalence. A prior report from the MiHEART study reported 49% plaque prevalence, according to editorialists led by Leslee J. Shaw, PhD, from the Icahn School of Medicine at Mount Sinai, New York, in an accompanying editorial – a much lower prevalence than detected by AI. In the current study, 63.3% of male participants Had detectable plaques using traditional human CTA analysis vs 81.5% using the AI-QCT method, with the same figures at 34.4% vs 61.9% respectively in women. “The much higher prevalence with AI-enabled coronary plaque analysis is concerning and in the absence of evidence reveals substantial disease overcalling,” the editorialists wrote. “The primary issue is whether AI-detected coronary plaque is even real when a study is visually normal. Importantly, overcalling is not only a technical limitation but also one which has detrimental clinical implications of overtreatment and potential patient anxiety related to this new pseudo disease.” Other findings The study found that the total plaque volume was 54 mm3 (Q1-Q3: 16-126 mm3), and this increased with age, the researchers noted. The amount of calcified plaque also increased with age. Male subjects had a greater median total plaque volume (80 mm3 [interquartile range: 31-181 mm3 ] vs 34 mm3 in women [interquartile range: 9-85 mm3]; P < 0.001), while 81.5% of male subjects had a total plaque volume ≥20 mm3 versus 61.9% of female subjects. No differences were observed according to race or ethnicity, the researchers said, highlighting the similarity between Hispanic and non-Hispanic subjects (53 mm3 [Q1-Q3: 14-119 mm3] vs 54 mm3 [Q1-Q3: 17-127 mm3]; P = 0.756). And younger individuals had a greater percentage of non-calcified plaques (89% of those in the age 40 to 44 and 45 to 49 year old groups, 86% of the 50 to 54 group, 78% of the 55 to 59s and 72% of those aged 60 to 64). ‘More realistic outlook’ “These data should be an important indicator for using quantitative coronary artery plaque volume in clinical practice, including in primary prevention settings,” the researchers concluded. “Despite the hope for AI-enabled coronary plaque analysis, recent research cast a less enthusiastic but perhaps a more realistic outlook for its integration into everyday imaging.” Source: Ichikawa K, Ronen S, Bishay R, et al. Coronary Plaque Volume in an Asymptomatic Population: Miami Heart Study at Baptist Health South Florida. JACC: Cardiovasc Imag; DOI: 10.1016/j.jcmg.2025.08.001. Shaw LJ, Blankstein R, Leipsic JA, et al. Clinical Integration of AI-Enabled Plaque Quantification to Improve Cardiovascular Risk Stratification. JACC: Advanc; DOI: 10.1016/j.jacadv.2025.09.001. Image Credit: Tierney – stock.adobe.com