AI-Driven Biomarker for Chronic Stress Identified Through Deep Learning Analysis of Routine Chest CT Scans
A groundbreaking study led by researchers at the Johns Hopkins University School of Medicine has identified the first visual biomarker of chronic stress detectable on standard medical imaging, marking a significant shift in how the physiological toll of psychological strain is measured and managed. Utilizing a sophisticated deep learning artificial intelligence model, the research team successfully quantified the size of the adrenal glands—the body’s primary stress-response organs—using existing chest CT scans. The findings, which demonstrate a direct correlation between adrenal volume and long-term cardiovascular risks, are scheduled for presentation at the annual meeting of the Radiological Society of North America (RSNA) next week.
Chronic stress has long been recognized as a silent driver of systemic disease, yet its measurement has remained notoriously elusive in clinical settings. Traditionally, clinicians have relied on patient questionnaires or single-point-in-time cortisol tests, both of which are subject to significant variability and bias. This new AI-driven approach offers a "long-term gauge" of stress, effectively visualizing the cumulative "wear and tear" on the body, known in medical terms as allostatic load. By leveraging the tens of millions of chest CT scans performed annually in the United States for other diagnostic purposes, this technology provides a secondary layer of health intelligence without the need for additional radiation or invasive testing.
The Biological Mechanism of Chronic Stress
To understand the significance of the Adrenal Volume Index (AVI) developed by the researchers, one must consider the role of the adrenal glands within the Hypothalamic-Pituitary-Adrenal (HPA) axis. When an individual experiences stress, the brain signals the adrenal glands to release hormones such as cortisol and adrenaline. While this "fight or flight" response is essential for survival in short bursts, chronic activation leads to structural changes in the glands.
Over time, the constant demand for hormone production can lead to adrenal hypertrophy—a physical enlargement of the glands. Until now, this enlargement was difficult to quantify precisely on routine imaging because the adrenal glands are small, irregularly shaped, and often overlooked when radiologists are focused on primary concerns like lung nodules or heart disease. The deep learning model developed by lead author Elena Ghotbi, M.D., a postdoctoral research fellow at Johns Hopkins, automates this process, providing a standardized measurement of adrenal volume that can be indexed against a patient’s height.
The health implications of this physiological change are profound. According to the American Psychological Association, chronic stress is a contributing factor to a litany of ailments, including clinical anxiety, insomnia, chronic muscle pain, and hypertension. By identifying a physical marker of this state, the research bridges the gap between psychological health and internal medicine, offering a tangible metric for a condition that has historically been treated as subjective.
Study Methodology and the MESA Cohort
The research was conducted using a robust and diverse dataset from the Multi-Ethnic Study of Atherosclerosis (MESA), a landmark longitudinal study supported by the National Heart, Lung, and Blood Institute. The team analyzed data from 2,842 participants with a mean age of 69.3 years, of whom 51% were women. This cohort was uniquely valuable because it provided a rare intersection of three data types: high-resolution chest CT imaging, validated psychological stress assessments, and biochemical markers of stress.
The researchers applied their AI model to the CT scans to calculate the Adrenal Volume Index (AVI), defined as adrenal volume in cubic centimeters divided by the square of the patient’s height in meters. This indexing ensures that the measurement is proportionate to the individual’s body size, allowing for a standardized comparison across a diverse population.
To validate the AVI as a true marker of stress, the team compared the AI-generated measurements against:
- Salivary Cortisol: Participants provided eight samples per day over a 48-hour period, allowing researchers to track peak levels and overall exposure.
- Allostatic Load: A composite score of physiological health including Body Mass Index (BMI), blood pressure, heart rate, and blood markers such as glucose, creatinine, and white blood cell counts.
- Psychosocial Indicators: Data from validated questionnaires measuring perceived stress and depression.
Chronology of the Research and AI Development
The development of this biomarker follows a decade-long trajectory of advancements in both medical imaging and machine learning. The MESA study itself has tracked participants for over 20 years, providing the "big data" necessary to correlate initial measurements with long-term health outcomes.
- Initial Data Collection (Early 2000s – Present): The MESA cohort was established to study the progression of subclinical cardiovascular disease. Over two decades, participants underwent multiple rounds of imaging and health assessments.
- AI Model Training (Recent Years): Dr. Ghotbi and her colleagues at Johns Hopkins trained a deep learning neural network to recognize the specific contours of the adrenal glands. This involved "teaching" the AI to distinguish the Y-shaped glands from surrounding fat, vascular structures, and other organs on chest CTs.
- Validation and Follow-up: Once the AI could accurately measure AVI, the researchers performed a retrospective analysis of the MESA data, looking back at up to 10 years of follow-up information to see how AVI predicted future health events.
- Presentation (Current): The culmination of this work is the presentation at the RSNA annual meeting, where the team will share the validated link between AVI and heart failure.
Key Findings: Linking Stress to Heart Failure
The results of the study indicate that the Adrenal Volume Index is not just a marker of past stress, but a predictor of future cardiovascular crisis. The research found that higher AVI values were significantly associated with greater overall cortisol exposure and higher peak cortisol levels. Furthermore, individuals who reported high levels of perceived stress on questionnaires were found to have higher AVI measurements than those who reported low stress.
Perhaps the most critical finding was the link between AVI and heart structure. Higher AVI was connected to a higher left ventricular mass index, a condition where the walls of the heart’s main pumping chamber thicken, often due to high blood pressure and chronic strain. The data revealed that for every 1 cm³/m² increase in AVI, the risk of heart failure and death increased significantly.
"With up to 10-year follow-up data on our participants, we were able to correlate AI-derived AVI with clinically meaningful and relevant outcomes," Dr. Ghotbi stated. She emphasized that this is the first imaging marker of chronic stress to show an independent impact on heart failure, even when accounting for other traditional risk factors.
Perspectives from the Research Team
The senior author of the study, Shadpour Demehri, M.D., a professor of radiology at Johns Hopkins, highlighted the practical advantages of using AI to "see" stress. He noted that while clinicians have long known that stress affects the body, they lacked a reliable way to quantify it during routine check-ups.
"For the first time, we can ‘see’ the long-term burden of stress inside the body, using a scan that patients already get every day in hospitals across the country," Dr. Demehri said. He pointed out that current methods, such as cortisol testing, are "cumbersome to obtain" and only offer a snapshot of a patient’s state at the moment the blood or saliva is drawn. The AVI, by contrast, represents the structural adaptation of the body to stress over months or years.
Teresa E. Seeman, Ph.D., a co-author and professor of epidemiology at UCLA, added a broader public health perspective. Having spent three decades studying how stress wears down the body, she described the work as a "true step forward in operationalizing the cumulative impact of stress on health." By linking a routine imaging feature with validated biological and psychological measures, the study provides a roadmap for integrating mental health metrics into physical medicine.
Broader Implications and Future of Clinical Practice
The implications of this research extend far beyond the field of radiology. If the Adrenal Volume Index is adopted as a standard metric, it could revolutionize cardiovascular risk stratification. Currently, doctors use tools like the Framingham Risk Score, which considers age, cholesterol, and blood pressure, to predict heart disease. Adding a validated stress biomarker like AVI could provide a more holistic view of a patient’s risk profile.
Preventive Care and Intervention
The ability to identify high-stress individuals through existing CT scans allows for "opportunistic screening." A patient undergoing a chest CT for a persistent cough might be identified as having a high AVI, prompting their primary care physician to intervene with stress-management strategies, blood pressure monitoring, or lifestyle changes before a major cardiovascular event occurs.
Addressing the Mental Health Crisis
As global rates of stress and burnout continue to rise, particularly in the wake of the COVID-19 pandemic, having a physical, quantifiable metric for stress could reduce the stigma associated with mental health conditions. When stress is shown to have a physical manifestation—similar to a clogged artery or a broken bone—it may encourage patients and providers to treat psychological well-being with the same urgency as physical health.
Application to Other Diseases
While the current study focused on heart failure, the researchers noted that this imaging biomarker could potentially be applied to a wide range of stress-related conditions. Chronic stress is known to exacerbate autoimmune disorders, accelerate cognitive decline, and worsen outcomes in cancer patients. Future research may explore how AVI correlates with these and other health challenges in middle-aged and older populations.
Analysis of the AI’s Role in Modern Medicine
The success of the Johns Hopkins team underscores the transformative role of AI in medical imaging. Radiologists are currently facing unprecedented workloads, with some reading hundreds of images per day. AI tools like the one developed for adrenal measurement do not replace the radiologist; rather, they perform high-precision tasks that would be too time-consuming or subtle for the human eye to perform consistently across thousands of cases.
By automating the segmentation and measurement of the adrenal glands, the AI extracts "hidden" data from scans that are already being performed. This maximizes the value of existing medical infrastructure. As healthcare systems move toward "value-based care," the ability to derive multiple diagnostic insights from a single test becomes increasingly important.
The presentation of this research at the RSNA meeting is expected to spark significant discussion among the global radiological community. As the first validated imaging marker of chronic stress, the Adrenal Volume Index stands as a testament to the power of combining long-term epidemiological data with cutting-edge artificial intelligence to solve some of medicine’s most persistent mysteries.
Related posts:
