Home / Knowledge base / Cardiology
Cardiology · echocardiography

A cardiac ultrasound reveals your biological age

Your chronological age is just a number. The heart ages by its own rules. Russian scientists trained a neural network to determine biological heart age from an ordinary ultrasound — echocardiography — with an accuracy of ±5 years. Let's unpack what exactly "ages" and how to measure it.

303patients in the sample ±5 yearsmodel accuracy 5echocardiography parameters ~8 minread
Biological heart age from echocardiography data
303
patients without chronic diseases in the RGNKTs training sample
±5years
mean absolute error in predicting heart age
0.78
R² — the share of age variation explained by the model from echocardiography data
8
echocardiography parameters "give away" the biological age of the heart

Biological heart age is a measurable quantity

In one 55-year-old, the heart works like that of a 40-year-old; in another, it already works like a 70-year-old's. The difference is driven by genetics, lifestyle, and accumulated strain — and until recently it was almost impossible to measure. Scientists from the N.I. Pirogov Russian National Research Medical University showed that an ordinary cardiac ultrasound can objectively assess this gap.

local_hospital
The study was conducted at the Russian Gerontology Research and Clinical Center (RGNKTs), N.I. Pirogov Russian National Research Medical University, Ministry of Health of Russia, Moscow. Authors: Kobelyatskaya A.A., Guvatova Z.G., Strazhesko I.D., Moskalev A.A. Published in the journal "Problems of Geroscience", 2023, No. 4: 215–218.
Chronological age
55
years on paper
Biological age
40–70
years — by echocardiography data

A heart's chronological and biological age can diverge by 10–15 years in either direction. Echocardiography makes it possible to measure this gap objectively for the first time.

What echocardiography shows — and why the heart "tells" its age

Echocardiography is an ultrasound examination of the heart. Unlike an ECG, which records only electrical signals, echocardiography provides a full "picture": how the walls move, how the valves open, how blood flows through the chambers of the heart. Some parameters change predictably with age — the heart "ages" along two interrelated scenarios.

foundation
Structural aging
  • The walls of the heart gradually thicken
  • The chambers change shape and volume
  • Myocardial hypertrophy appears
  • Captured by IVS, LV_PW, RWT
water_drop
Functional aging
  • The myocardium relaxes more slowly
  • Ventricular filling worsens
  • Diastolic dysfunction develops
  • Reflected in E/A and cardiac output

"Echocardiography allows morphological and functional changes of the heart to be assessed simultaneously: wall dimensions, chamber volumes, blood-flow velocities. Some of these measures correlate strongly with age" — and it is on this principle that the model is built.

The study: 303 patients, 48 parameters — and rigorous selection

The work drew on echocardiography data from 303 RGNKTs patients without chronic diseases: aged 23 to 82 years in women and 25 to 72 years in men. All parameters in absolute units (centimeters, liters) were normalized to the patient's height — to eliminate the influence of body build.

analytics
Parameter selection method: Spearman correlation analysis. Of the 48 measured indicators, those correlating with age above |0.55| were selected — a strong association. Passing the threshold were 15 parameters, of which 8 are part of the standard echocardiography protocol.

An important nuance: the strength of these parameters' correlation with age differs between men and women, though the direction of the association holds for both sexes. So the scientists built two independent models — one for men and one for women — and then combined the best variants into final composite models.

8 echocardiography parameters that "give away" biological age

Six of them are direct echocardiography measures; two (H_RWT and L_E/A) are derived, computed automatically from the primary ones. The final neural-network model includes 5 of the 6 direct parameters (without LV_EDV) and both derived ones.

E/A
E/A ratio
Early/late LV filling velocities. In the young >1, it declines with age
IVS
Interventricular septum
Thickness of the wall between the ventricles (cm/height). Grows with age
LV_PW
LV posterior wall
Thickness of the left ventricle's posterior wall (cm/height). A marker of remodeling
LV_CO
Cardiac output
Volume of blood pumped by the heart per minute (L/min). Changes with age
LV_EDV
LV end-diastolic volume
End-diastolic volume: how much blood is in the LV before ejection
RWT
Relative wall thickness
Relative thickness of the LV walls. An integral marker of remodeling
H_RWT
RWT ≥ 0.42
Derived: a binary indicator of concentric hypertrophy
L_E/A
E/A < 1
Derived: a marker of impaired diastolic function

How the key parameters change with age

E/A
declines with age
IVS
grows with age
RWT
grows with age
LV_PW
grows with age
LV_CO
declines with age
LV_EDV
declines with age

All these indicators passed the age-association threshold of |ρ| > 0.55 (a strong Spearman correlation). The specific coefficients are not reported in the primary source — what matters is the direction of change with age.

Why E/A is the most powerful predictor of age

The E/A ratio is the ratio of peak blood-flow velocities in the two phases of diastole (relaxation) of the heart. When you are young, the heart muscle is elastic: the early filling phase (E) is fast and forceful, and the E/A ratio is well above 1. With age the muscle loses elasticity and relaxes worse: the early phase weakens, and the atria have to work harder (the late phase A grows). When E/A falls below 1, that is diastolic dysfunction — one of the early signs of an "aging" heart.

Why the walls thicken with age

The IVS (interventricular septum thickness) and LV_PW (posterior wall) parameters reflect how the myocardium adapts. Under chronic strain — pressure, metabolic changes — the walls gradually thicken. When relative wall thickness (RWT) is ≥ 0.42, one speaks of concentric remodeling or myocardial hypertrophy.

How the neural network works — the FCNN architecture

The scientists built a fully connected neural network (FCNN — Fully Connected Neural Network) with 10 hidden layers on the Keras platform. The data were split into training and test sets.

cardiology 5 echocardiography parameters
arrow_forward
straighten Normalization to height
arrow_forward
wc Model choice ♂/♀
arrow_forward
neurology FCNN, 10 layers
arrow_forward
schedule Biological heart age
data_object
The model's full set of input data: LV_CO (L/min), E/A (velocity ratio), RWT (relative wall thickness), IVS (septum thickness, cm/height), LV_PW (LV posterior wall, cm/height), plus two binary derived features — H_RWT (RWT ≥ 0.42) and L_E/A (E/A < 1) — and the patient's height (cm).

The model was trained in three modes: the whole sample, men only, and women only. The best results for each sex were combined into a final composite model. This ensemble approach made it possible to squeeze the most out of a relatively small sample (303 people).

lightbulb
Why is normalization to height important? A person 190 cm tall has a physically larger heart than a person 160 cm tall — and that says nothing about their age. Normalization removes this "noise" and makes the parameters comparable.

Results: accuracy within ±5 years

The final models for men and women showed comparable results. The key quality metrics: MAE (mean absolute error), R² (share of explained variation), and ε-accuracy (the share of predictions within ±10 years of the true age).

Men ♂
±4.9 years mean absolute error (MAE)
R² = 0.78 78% of age variation explained by the model
88% of predictions fell within the ±10-year range
Women ♀
±5.1 years mean absolute error (MAE)
R² = 0.77 77% of age variation explained by the model
89% of predictions fell within the ±10-year range

R² = 0.77–0.78 means: knowing only the echocardiography parameters, the neural network explains almost 78% of the spread of age in the sample — with no other data about the patient. This is a strong result, given that biological age depends on dozens of factors beyond the state of the heart.

balance
For context: earlier AI models based on the ECG (the 2025 EHRA study on 500,000+ records) determined the biological age of the heart and predicted the risk of death. The echocardiography-based model adds a structural dimension — information about the shape and walls of the heart that the ECG does not see. For more on the related method, see the article on how AI determines heart age from an ECG.

What this means for you in practice

Echocardiography is a standard test that a cardiologist orders for heart complaints, shortness of breath, or high blood pressure. If you already have results, you already have the data to assess your heart's biological age. And while you don't yet have it, it's worth assessing your 10-year cardiovascular risk with the QRISK calculator.

01
Check for E/A in your echocardiography report

This is the most important parameter. If E/A < 1, that's a signal of impaired diastolic function, typical of an "aged" heart. Discuss with your cardiologist what lies behind it.

02
Pay attention to RWT

RWT ≥ 0.42 means concentric remodeling — hypertrophy of the heart walls. On its own it's not a diagnosis, but it's a reason to keep blood pressure and strain in check.

03
Aerobic exercise improves diastolic function

Regular cardio is one of the few proven ways to improve diastolic function and "rejuvenate" the heart muscle by several years on echocardiography.

04
Combine echocardiography and ECG for the full picture

The ECG sees electrical activity, echocardiography sees structure and mechanics. Together they give the most complete portrait of your heart's biological age.

Echocardiography is not just a "picture" of the heart

Five parameters — and a neural network will tell you, to within ±5 years, how much younger or older your heart is than your passport. Biological heart age is a measurable — and therefore manageable — indicator.

monitor_heart Article · cardiology AI determines heart age from an ECG How a neural network reads an ordinary ECG and why a 7-year gap between biological and chronological heart age changes the prognosis — from an analysis of 500,000+ records. Read the article

Related materials

The rest is subscription-only

This section is available on a higher plan. Get access to unlock all materials.