Vascular aging shows up as a gradual hardening: the pulse wave travels faster, reflected waves return earlier, and the systolic load on the heart grows. Doppler analysis of the waveform in the carotid artery yields physiologically meaningful information about resistance, elasticity, and wave reflection. Nearby in the archive sit studies that also extract diagnostic features from signals: the mathematics of arterial tortuosity, kettle noise for quantum sensors, a neural network for gravitational lenses, MRI discrimination of water in tissues. All of them hunt for hidden patterns in data, but this paper stands apart: its closest neighbor along the meaning vector is far away (0.487), and nothing is right next to it. The map's thematic group, quantum algorithms, error correction, quantum information, does not overlap with vascular physiology. This underscores that the work tackles a specific problem, estimating age from a reflected wave, without leaning on neighboring methods.
197 participants (123 men, 74 women) aged 20 to 67 underwent screening: normal blood pressure (systolic ≤ 140 mm Hg), body mass index ≤ 30 kg/m², no cardiovascular medications and no chronic diseases. Screening relied on self-report and resting blood pressure, not on laboratory panels. The Doppler signal from the left common carotid artery was recorded with a portable wireless sensor (2.0 MHz, 50° angle), digitized at 10 kHz, and averaged over 30 artifact-free cycles per person. From the velocity envelope, eight features were extracted, including peak systolic velocity S1, reflected systolic velocity S2, the incisura I, diastolic velocities D and d-min, plus three dimensionless indices: RI (resistance), VRI (relative magnitude of the reflected systolic component), and VEI (elastic recoil). The link to age was assessed with Pearson correlation with confidence intervals, multiple regression adjusted for sex and exercise habits, and a random forest (500 trees) with repeated 10-fold cross-validation (10 repeats), permutation importance on held-out test folds, and a bootstrap of 1000 resamples. The Benjamini–Hochberg correction was applied to eight pre-specified robustness tests.
Age correlated most strongly with VRI: r = 0.738 (95% CI [0.667; 0.796], p < 0.001). Negative correlations with age were found for S1 (r = -0.627), RI (r = -0.562), VEI (r = -0.435), and D (r = -0.290); S2 had a weak positive correlation (r = 0.201), while I and d-min were not significant. After adjusting for sex and exercise habits, the partial correlation of VRI with age rose to r = 0.798. In the random forest, VRI received the highest impurity importance (0.536), followed by S1 (0.218), with the remaining features contributing less than 0.06. Permutation importance on held-out folds confirmed VRI's lead: the mean drop in R² was 0.765 ± 0.178 versus 0.110 ± 0.060 for S1, roughly seven times greater. In the bootstrap of 1000 resamples, VRI ranked first in 85.3% of cases, and its mean importance was 0.516 (95% CI [0.213; 0.687]). VRI added significant information to the other seven features: R² increased from 0.593 to 0.646 (partial F = 27.95, p < 0.001), and VRI alone explained 0.545 of the variance in age. Exercise habits were associated with lower VRI after adjustment for age only (B = -0.043, p = 0.018), but after further adjustment for sex the effect weakened and became non-significant (B = -0.026, p = 0.098). Sex remained significant after adjustment for age and height (p < 0.001), although height alone was borderline (p = 0.097).
The work differs from its archive neighbors in that it does not build a complex model or a new sensor, but tests a single physiologically interpretable feature for robustness to adjustments and alternative importance estimation methods. Neighboring studies extract features from noise or images, but do not link them to age-related vascular physiology. Here, VRI showed the strongest and most stable association with age, making it a candidate biomarker of vascular aging. However, without comparison to reference measures of arterial stiffness (carotid-femoral pulse wave velocity, augmentation index, intima-media thickness), equivalence cannot be claimed. This is not validation, but hypothesis generation.
The topic could move toward directly comparing VRI with reference measures of arterial stiffness in an independent cohort with a wider age range, especially above 67. If the link holds, a portable Doppler sensor could replace bulky equipment for assessing vascular age in the field or in a general practitioner's office. It would also be promising to study how VRI changes with physical exercise, drug therapy, and in patients with cardiovascular disease.
The results could influence non-invasive diagnosis of vascular aging, preventive cardiology, and sports medicine, where a quick assessment of arterial condition matters. The portability of the method opens the door to mass screening outside specialized laboratories.
Conduct a validation study with simultaneous measurement of VRI and carotid-femoral pulse wave velocity in the same group. Include participants over 67 and expand the sample to test interactions with age and sex.
In vascular physiology, it remains unresolved how well chronological age reflects biological aging of the arteries. VRI could become a bridge between the Doppler waveform and reference stiffness measures, but a direct comparison in a single cohort is still lacking. Also open is the question of causality between physical activity, sex, and the reflected wave: the cross-sectional design cannot establish the direction of effects.
The gap between the strongest and weakest feature in the random forest was almost 20-fold: VRI received an importance of 0.536, while d-min got just 0.025. This shows how much more informative the reflected wave is compared to the diastolic minimum.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.