AI in Cardiac CT

Is AI already part of my cardiac imaging workflow?

It probably is. You just might not realise it. AI has been quietly embedded in many radiology platforms for years, handling image segmentation and generating curved multiplanar reformats in the background without revealing itself. 

Why is there such urgency around AI in cardiac imaging?

Here’s the ugly truth. Training a radiologist takes 11 years, and by 2050, roughly a quarter of the world’s population will be over 60. Demand for cardiac imaging is already outpacing the workforce, and that gap is only going to widen. In this regard, AI isn’t a luxury, it’s a practical response to a problem that isn’t going away. 

What does a fully automated cardiac CT report look like in practice?

In a live demonstration during the webinar, one platform produced a complete CCTA report with plaque analysis, calcium score, and CT-FFR included all within eight to ten minutes of the scan being performed. The clinician reviews the output, makes any edits, and approves it. It’s less of a replacement of the reporting process and more a very fast first draft that’s already done most of the heavy lifting.

If an AI tool makes a diagnostic error, who is responsible?

The question that almost all of us are thinking about. And the answer that AI companies would prefer not to advertise. Under most current legal and regulatory frameworks, the clinician retains primary responsibility for patient care, full stop. These tools are classified as decision support, not autonomous decision-makers. The AI suggests, the clinician decides.

Can AI reliably automate TAVI CT planning?

Reporting a TAVI scan is one of the most time-consuming tasks in cardiac imaging, regularly taking 30 to 40 minutes of careful manual measurement per case. AI platforms can reduce that down to a couple of minutes end-to-end, with less than one minute of clinician screen time. In terms of accuracy, they have been validated against expert measurements across more than 1,000 real-world patients and exceeded 98%.

How can cash-strapped hospitals afford expensive AI tools?

It’s a fair question, and one that was put directly to a company presenting its technology. Their answer focused on efficiency. If AI can reduce the time a clinician spends reviewing a case from several minutes to less than a minute, the same team can manage a much larger workload. In that sense, the productivity gains can help offset the cost of the technology. A second point was that hospitals may not need to carry the full financial burden alone. Other stakeholders, such as device manufacturers and industry partners who benefit from these analyses, may also have a role in supporting adoption. 

What can AI detect in a cardiac CT that the human eye cannot?

Two-thirds of major cardiac events after CCTA occur in patients without obstructive disease, a group often reassured and discharged to primary care. In a 40,000-patient UK cohort, individuals without obstructive CAD accounted for 66.3% of all major adverse cardiac events and 63.7% of cardiac deaths over a median 2.7 years.

The hidden driver of much of this risk is coronary inflammation. AI-enabled analysis of standard CCTA can now quantify coronary inflammation indirectly, by reading subtle changes in the surrounding perivascular fat. AI can transform a routine CCTA into a map of active coronary disease biology, revealing inflamed, high-risk arteries in scans that may look “low risk” on conventional visual review

Who should benefit when patient data is used to build commercial AI tools?

One view put forward in the webinar was that value generated by AI tools built on NHS patient data ought to find its way back into the NHS, potentially through revenue-sharing arrangements. Whether that ever becomes policy is another matter, but it’s a question the field is going to have to navigate.

References

• Rajani R. AI in Cardiac CT. Medmastery Free Webinar
• Rajani R. Cardiac CT Essentials. Medmastery
• Elshahati N. AI Scribes for Clinical Practice. LITFL
• Wiesbauer F. ChatGPT Essentials for Clinicians. Medmastery
• Guilleminot S. AI in Healthcare. LITFL