AI Uncovers Hidden Heart Risks in Routine Scans
A new artificial intelligence tool is revealing hidden heart risks that are often missed in standard chest scans. The AI analyzes existing CT scans, offering a proactive approach to heart health and potentially saving lives through early detection.
AI Detects Silent Heart Warnings
Heart disease remains a leading cause of death in the United States, claiming over 700,000 lives in 2022. Millions undergo chest computed tomography (CT) scans annually for unrelated reasons; these scans can also provide critical insights into heart health, highlighting calcium deposits that signal potential coronary artery issues.
An algorithm, named AI-CAC, was developed by Dr. Hugo Aerts and his team at Mass General Brigham, in collaboration with the U.S. Department of Veterans Affairs. The AI-CAC tool transforms overlooked data from CT scans into an early warning system for heart problems.
The presence of coronary artery calcium indicates past plaque damage. Any calcium detected is a red flag, and a high score can dramatically increase the risk of heart attacks. Despite this, standard chest CT scans often overlook the heart, leading to late-stage diagnoses.
How AI Improves Detection
The AI-CAC system utilizes deep learning, a method that allows software to learn patterns by examining numerous labeled images. Developers trained the network using scans from 98 VA hospitals, representing various scanner types and patient body sizes. This training allows the AI to identify calcium deposits with high accuracy.
“Millions of chest CT scans are taken each year, often in healthy people.”
—Dr. Aerts
The AI correctly identified calcium deposits with 89% accuracy. Its speed is also a major advantage, taking seconds compared to the minutes required by human experts. Currently, the Centers for Disease Control and Prevention (CDC) reports that heart disease accounts for about 1 in every 4 deaths in the United States (CDC 2024).
Impact on Patient Care
Patients with high AI-CAC scores faced a significantly increased risk of mortality. Independent cardiologists reviewed a sample of high-score scans and agreed that most patients would benefit from cholesterol-lowering medications, highlighting the model’s clinical importance. This proactive approach can lead to earlier interventions, potentially reducing the severity of heart disease.
The project’s strength lies in its large and diverse dataset, a contrast to previous calcium-AI efforts that used gated scans from a single vendor. The dataset originated exclusively from veterans, but expanding validation across diverse demographics and scanner models is underway.
Dr. Raffi Hagopian, the study’s lead author, sees the tool as an ideal initial step, noting the availability of existing chest CT scans suitable for opportunistic screening within VA imaging archives. He also envisions a shift towards proactive disease prevention, potentially reducing heart attacks and improving patient-clinician decision-making.
Addressing Future Challenges
Automating calcium detection can relieve radiologists of a tedious task and flag high-risk patients. However, there are concerns regarding false positives and data security. Health systems must also determine liability in case of algorithm errors.
Future steps involve testing AI-CAC in community hospitals, monitoring the impact of early statin prescriptions, and integrating the calcium score into electronic health records. Researchers are exploring expanding the AI’s capabilities to measure aortic and valvular calcium, enhancing its scope without additional radiation exposure.
AI-CAC offers a path to use existing knowledge to benefit every patient who has undergone a chest CT scan. It is published in the journal Nejm who.
