AI Spots Stress in CT Scans

FEATURED RESEARCH

The First Evidence That Stress Can Be Measured Directly on CT Imaging

Chronic stress is one of those things everyone feels but medicine rarely measures this well. A new study from Johns Hopkins finally gives us a potential physical marker.

Using a deep learning model, the team found that the adrenal glands enlarge in people living with long-term stress and that this signal shows up on everyday chest CTs already used across hospitals.

How the model works: Chest CTs are done tens of millions of times a year, and the adrenal glands sit right there next to the kidneys. The team trained an AI system to segment these glands and calculate their volume automatically.

The idea is simple: cortisol spikes come and go, but the long grind of stress pushes the adrenal system over time. That growth becomes a long-term footprint.

The dataset came from nearly 3,000 adults in the Multi-Ethnic Study of Atherosclerosis.

This group had chest CTs, detailed stress questionnaires, cortisol measurements across the day, and markers of allostatic load (the wear-and-tear that builds up from chronic stress).

It’s one of the only cohorts in the world with all these variables in one place.

What they found: Higher adrenal volume matched higher cortisol, higher allostatic load, and higher perceived stress. It also predicted health outcomes up to a decade later.

Each small increase in adrenal size is linked to a greater risk of heart failure and mortality. The signal held even after adjusting for traditional risk factors.

Why it matters: This is the first imaging biomarker of chronic stress that’s been validated against biology and long-term outcomes! And because the signal lives inside a scan people already get, it doesn’t add radiation, cost, or extra testing.

If stress leaves a footprint, giving clinicians a way to see it is a major step toward catching problems earlier and maybe treating people more humanely.

For more details: Full Article 

Brain Booster

Select the right answer! (See explanation below and source)

What Caught My Eye

AI SCANDAL

Deloitte Under Fire After A Second Public Report Shows Signs of AI-generated Inaccuracies

Newfoundland and Labrador (Canada) just joined the growing list of governments discovering AI-generated errors inside major policy documents. This time, it was a $1.6 million Health Human Resources Plan from Deloitte that included citations to research papers that don’t appear to exist. 

It’s the second government report in the province that questioned for fabricated sources in the span of a few weeks.

The report was meant to guide long-term planning for a healthcare system already struggling with staff shortages. Instead, researchers named in the citations publicly said the papers were false and likely produced by an AI system.

One cited expert confirmed her team never produced the cost-effectiveness analysis the report claimed. Another said a paper attributed to her simply did not exist.

The scandal lands at an uncomfortable moment for healthcare leaders who are being told AI can streamline planning, improve forecasting, and support clinical decisions. Trust is the foundation of that pitch. When governments discover phantom studies in flagship reports, confidence erodes fast.

Unfortunately, this follows from last month, when Deloitte again made international headlines after its firm in Australia produced a report for the Australian government that contained apparent AI-generated errors…

For more details: Full Article

Top Funded Startups

Byte-Sized Break

📢 Other Happenings in Healthcare AI

• Patients are turning to AI tools like Sheer Health and Counterforce Health to fight rising health insurance denials, creating an AI vs. AI dynamic as states move to regulate insurer use of AI and ensure human oversight in medical decision-making. [Link]

• Canadian mental health experts are urging parents to think carefully before buying AI-enabled toys that track kids’ speech, moods, and behaviour. The concerns sit at the mix of data privacy, emotional attachment to bots, and the idea of outsourcing social interaction to devices marketed as “smart” companions. [Link]

• Penn Medicine's AI system, coordn8, has saved 8,500 staff hours by automating fax filing and digital consent, tripling processing speed and cutting a week off new patient intake time. [Link]

Have a Great Weekend!

Trivia Answer: B) Hippocampus

The hippocampus plays a key role in memory and regulating the body’s stress response. Chronic stress and elevated cortisol can lead to hippocampal shrinkage, which may affect learning and emotional regulation. This connection is one reason why long-term stress can have cognitive effects.