Longevity & AgingResearch PaperOpen Access

Your Thymus Keeps Protecting You Into Old Age — and AI Can Now Measure How Well

A deep learning system measuring thymic health from CT scans reveals that better thymic function predicts longer life and lower cancer and heart disease risk.

Sunday, May 24, 2026 16 views
Published in Nature
A translucent glowing thymus gland centered in a human chest, surrounded by branching immune cell networks, cool blue medical tone

Summary

Researchers developed an AI system to quantify thymic health from routine CT scans and applied it to over 27,000 adults across two large cohorts. They found that individuals with higher thymic health lived longer, had lower rates of lung cancer incidence, and experienced less cardiovascular mortality over 12 years of follow-up. Thymic health also tracked with systemic inflammation and metabolic markers, and was measurably worse in smokers, obese individuals, and those with low physical activity. These findings challenge the long-held belief that the thymus becomes irrelevant after childhood, repositioning it as an ongoing regulator of immune-mediated ageing that may be targetable through lifestyle changes or regenerative therapies.

Deep Dive Audio
0:00--:--

Detailed Summary

For decades, the thymus — the organ responsible for producing and diversifying T cells — was considered functionally obsolete in adults. The prevailing view held that once a diverse T cell repertoire is established in childhood, peripheral maintenance is sufficient. This study, published in Nature, directly challenges that assumption with large-scale epidemiological evidence.

The research team built a deep learning pipeline trained on 5,674 CT scans to automatically localize the thymus and quantify its composition as a continuous proxy for thymic functionality, termed 'thymic health.' The system was then validated on 27,612 individuals from two independent prospective cohorts: the National Lung Screening Trial (NLST, n=25,031) and the Framingham Heart Study (FHS, n=2,581). Participants were stratified into low (bottom 25%), average (middle 50%), and high (top 25%) thymic health categories.

In the NLST, higher thymic health was consistently associated with lower all-cause mortality, reduced lung cancer incidence, and lower cardiovascular mortality over 12 years of follow-up, after adjustment for age, sex, smoking status, and comorbidities. These findings were independently replicated in the FHS, where higher thymic health was significantly associated with reduced cardiovascular mortality. The convergence of results across two distinct cohorts with different designs substantially strengthens the causal argument.

Thymic health also correlated with systemic inflammation and metabolic dysregulation, suggesting mechanistic pathways through which thymic decline may accelerate age-related disease. Critically, several determinants of thymic health were modifiable: smoking, obesity, and physical inactivity were each associated with lower thymic health, while female sex and younger age predicted higher values — consistent with known biology of thymic involution.

These findings reframe thymic health as a measurable, potentially modifiable biomarker of immunological ageing. They open clinical avenues for thymus-targeting interventions — from lifestyle modification and metabolic optimization to experimental regenerative strategies — as tools to promote healthy ageing. The ability to assess thymic health non-invasively from routine CT scans, which are already widely used in lung cancer screening, makes population-level application feasible.

Key Findings

  • AI-quantified thymic health from CT scans predicted all-cause mortality over 12 years in 25,031 adults.
  • Higher thymic health was independently linked to lower lung cancer incidence and cardiovascular mortality.
  • Results replicated in the Framingham Heart Study cohort of 2,581 participants.
  • Smoking, obesity, and physical inactivity were each associated with worse thymic health.
  • Thymic health correlated with systemic inflammation and metabolic dysregulation markers.

Methodology

A self-supervised deep learning model was trained on 5,674 CT scans to quantify thymic composition and applied to 27,612 adults in two prospective cohorts (NLST and FHS). Outcomes including all-cause mortality, lung cancer incidence, and cardiovascular mortality were assessed over up to 12 years, with Cox regression models adjusted for age, sex, smoking, and comorbidities.

Study Limitations

The study is observational and cannot establish causation between thymic health and disease outcomes. CT-based thymic composition is an indirect proxy for T cell output and may not fully capture functional immunity. Cohorts were predominantly older adults, limiting generalizability to younger populations.

Enjoyed this summary?

Get the latest longevity research delivered to your inbox every week.