Jack D. Hidary is the CEO of SandboxAQ, which focuses on enterprise SaaS solutions at the convergence of AI and Quantum tech.
Cardiovascular disease is the number one killer across most of the world. Each year, there are 400 million cases of ischemia, heart failure and arrhythmia globally. Heart disease is the leading cause of death in the U.S. which, combined with stroke, translates to an economic burden of more than $254 billion. Costs from cardiovascular diseases are projected to hit roughly $2 trillion by 2050.
The standard diagnostic tool most widely used to monitor heart activity and detect cardiac issues—the electrocardiogram (EKG)—has vastly improved cardiac care but it has limitations. Most notably, when measured by EKG, the heart’s electrical signals may be distorted as they pass through the body. This can complicate diagnosis and lead to inaccurate and potentially dangerous conclusions. EKG is more than 150 years old and thus ripe for disruption by more sophisticated technologies.
Thanks to breakthroughs in artificial intelligence and the availability of more powerful sensors, a new class of highly advanced, more accurate magnetocardiography (MCG) tools are drawing positive attention in the medical community. MCGs measure cardiac activity by analyzing changes in the magnetic field emanating from the heart. Unlike EKGs, these signals are not distorted by human tissue, so the diagnostic data are delivered at a much higher fidelity.
Although MCG-based devices have existed for decades, their adoption and use has been limited to larger medical facilities due to their high cost, large size and complexity. For example, early MCG devices based on SQUIDs (superconducting quantum interference devices), require a large, dedicated room and need specialized electromagnetic shielding and cryogenic cooling systems to operate, which increases costs. These factors prevented them from scaling beyond a handful of large medical centers.
The next generation of MCG devices, however, are much smaller and self-contained, able to be installed next to a hospital bed or brought to patients on a cart for point-of-care use. They operate at room temperature with low power requirements and don’t need complex cooling or shielding systems. This significantly opens up MCG’s potential use in the ER, ambulances, field clinics and beyond. It also makes it more cost-effective for smaller facilities to own and operate and, thus, less costly for patients and insurers. Finally, it’s also more comfortable for patients, who need not endure food or water restrictions and injectable dyes prior to imaging, or suffer long, uncomfortable procedures inside a confined space.
How realistic is AI in healthcare?
The use of AI in healthcare is experiencing a boon in both drug discovery (a topic for another column) and medical devices. The AI-enhanced diagnostic market is set to grow from $16.3B in 2023 to $71.2B in 2027. Next-generation MCGs have the potential to elevate patient care by filling the gap between EKGs and biomarker tests (which measure cardiac enzyme levels and take hours for lab analysis), and costly and complex MRI, PET, CT and invasive angiography procedures.
AI is a critical factor driving these next-gen MCG devices. First, we can train AI models to filter out external interference that exists within a dynamic hospital environment—including electromagnetic interference from lights, elevators and other hospital equipment—eliminating the need for physical shielding and significantly reducing device cost and size. More importantly, AI is needed to convert the heart’s magnetic signals into visualizations that depict the patient’s cardiac electrical signals and ultimately, from signals to actionable insights that physicians can use.
As with other AI tools, there are inherent risks with implementing AI-powered MCG devices. Training is incredibly important—both for the AI models and the personnel using these devices. Large quantitative models trained on the mathematical underpinnings of anatomy, biology, physiology and pharmacology are inherently less prone to hallucinations than large language models trained on medical text, but their training could be impacted if the available data isn’t as sufficient, representative or diverse as required. Misinterpretations caused by under-trained personnel can lead to false positives or negatives, unnecessary interventions, or missed diagnoses.
These risks can be overcome by device manufacturers working closely with clinicians to develop intuitive user interfaces, or training AI models in conjunction with other diagnostic modalities and comparing the results.
At first, medical professionals might not trust the AI-generated MCG interpretations and could continue putting patients through more traditional diagnostic tests (impacting patient care, ER efficiency and cost of care). On the flip side, busy professionals may become overly reliant on AI-generated insights instead of leveraging their clinical expertise when making diagnoses. Implementing AI systems at scale is about change management as much as it is about technology, and educating clinicians on how to work alongside AI is half the battle.
Another downside to implementing next-generation MCG devices is cost. Large facilities that have already spent millions on MRI, PET, CT and other tools might seek to maximize the ROI on these devices before replacing them or investing in new ones. Smaller or more remote hospitals might not have the budget to invest in MCG, or they might not have the right personnel on staff (e.g., interventional cardiologists) to treat the conditions once diagnosed.
Speed and accuracy of cardiac diagnosis are critical to improve patient care and survivability during a cardiac event. For example, in the U.S., chest pain is currently one of the top reasons for patient ER visits. According to National Hospital Ambulatory Medical Care Survey data, there are more than 8 million chest pain cases annually in U.S. emergency rooms—with only a fraction caused by heart attacks. Many cases could be moved immediately out of the ER if a quick and accurate test was available on-site or even in the ambulance en route to the ER. Many patients being evaluated for a suspected heart attack are subsequently admitted to the hospital for a short stay, which bears an annual cost of $3 billion. If clinicians have tools that can help them more quickly differentiate benign chest pain from an acute heart attack, we could reduce costs and get patients triaged more accurately and treated faster before things get worse.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
