Explainable Machine-Learning Analysis of Carotid Doppler Waveform Biomarkers for Chronological-Age Estimation

Q: What are the key findings of this study?

Carotid Doppler waveform features estimated chronological age with R² = 0.609 and MAE 8.27 years, while Velocity Reflection Index correlated strongly with age.

Q: What is the clinical evidence from this study?

Study design: Cross-Sectional. Population: Chronological aging / Vascular aging in healthy adults (n=227). Intervention: Carotid Doppler waveform biomarkers (8 features including Velocity Reflection Index [VRI]) vs. 7-feature subset without Velocity Reflection Index (VRI) in ablation analysis. Primary outcome: Chronological age estimation accuracy and correlation of Velocity Reflection Index (VRI) with age (r = 0.739 (VRI correlation with age); RF model R² = 0.609 ± 0.140, MAE = 8.27 ± 1.40 years, p=<0.001).

Q: What does this research mean for the field?

Carotid Doppler waveform features can estimate chronological age with moderate-to-high accuracy (R² = 0.609, MAE = 8.27 years), with the Velocity Reflection Index showing the strongest individual correlation with age. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Q: What question did this study set out to answer?

To describe an explainable biomarker-discovery framework linking signal processing, explainable machine learning, and physiological interpretation using carotid Doppler waveforms to estimate chronological age.

Azran Azhim

doi:10.21203/rs.3.rs-11505581/v1

November 1, 2026Open Access

Explainable Machine-Learning Analysis of Carotid Doppler Waveform Biomarkers for Chronological-Age Estimation

View Full Paper

Ask AI

Bookmark

Share

Key result

Carotid Doppler waveform features estimate chronological age within ~8 years.

r = 0.739 (VRI correlation with age); RF model R² = 0.609 ± 0.140, MAE = 8.27 ± 1.40 years
P<0.001
n=227

Why the study?

Physiological sensing and machine learning are increasingly combined for health assessment, but many systems emphasize predictive accuracy over physiologically grounded explanation.

Does an explainable machine-learning framework applied to carotid Doppler waveform biomarkers accurately estimate chronological age and reliably quantify individual feature contributions in healthy adults?

Key Points

To describe an explainable biomarker-discovery framework linking signal processing, explainable machine learning, and physiological interpretation using carotid Doppler waveforms to estimate chronological age.
Assessed 227 healthy adults aged 20–89 years (136 males, 91 females) who underwent blood-flow-velocity waveform recordings from the left common carotid artery using a portable Doppler ultrasound system.
Extracted eight waveform landmark and derived features and related them to chronological age using Pearson correlation and Random Forest regression (500 trees) with repeated 10-fold cross-validation.
Quantified feature importance using Gini importance and permutation methods (in-sample and held-out), combined with ablation of the top-ranked feature to test unique predictive contributions.
Velocity Reflection Index (VRI) correlated most strongly with age (r = 0.739, p < 0.001), while the primary systolic peak (S1) showed a strong negative association (r = -0.661, p < 0.001).
The Random Forest regression model estimated chronological age with an MAE of 8.27 ± 1.40 years and an R² of 0.609 ± 0.140.
VRI dominated Gini importance (50.4%) and held-out permutation importance, but retraining after removing VRI reduced R² by only 0.043 (from 0.609 to 0.566) due to strong correlation with other retained landmarks (r = -0.64).

Study Design

Type

Cross-Sectional (n=227)

Multicenter

No

Structured PICO

Does an explainable machine-learning framework applied to carotid Doppler waveform biomarkers accurately estimate chronological age and reliably quantify individual feature contributions in healthy adults?

P

Population

227 healthy Japanese adults aged 20–89 years (mean age 40.2 years, 40.1% female) free of known cardiovascular disease, arrhythmia, diabetes, or treated hyperlipidemia evaluated cross-sectionally via portable carotid Doppler ultrasonography.

E

Exposure

Portable carotid Doppler ultrasonography (left common carotid artery at 2 MHz, 50° insonation angle) and an Explainable Biomarker-Discovery Framework utilizing Random Forest regression (500 trees, repeated 10-fold cross-validation) on eight waveform features (S1, S2, I, D, d-min, RI, VRI, VEI), evaluated via Gini importance, permutation importance, and feature ablation.

C

Comparator

Feature ablation subsets (removing Velocity Reflection Index [VRI], S1+S2 only, VRI only, S1+S2+VRI) and comparison across three feature-importance methods (Gini importance, in-sample permutation importance, held-out cross-validation permutation importance).

O

Outcome

Chronological-age estimation accuracy (measured by Mean Absolute Error [MAE], Root-Mean-Square Error [RMSE], and coefficient of determination [R²]) and the unique variance contribution of the Velocity Reflection Index (VRI) under feature ablation.surrogate

Main Result

Effect estimate: r = 0.739 (VRI correlation with age); RF model R² = 0.609 ± 0.140, MAE = 8.27 ± 1.40 years

Absolute Event Rate: 0.609% vs 0.566%

p-value: p=<0.001

Carotid Doppler waveforms carry substantial age-related hemodynamic information (estimating chronological age with R² = 0.609), but standard machine-learning importance metrics severely overstate the unique contribution of correlated biomarkers like the Velocity Reflection Index unless paired with feature ablation.

Limitations

Construct validity is limited as the output is a chronological-age estimate rather than a validated vascular-age biomarker (no PWV, intima-media thickness, or clinical outcome data).
Instance-level and counterfactual explanations were not computed.
Blood pressure, BMI, and body-composition data were incomplete and missingness was significantly age-related (younger participants had complete data); smoking and medication status were not recorded.
Repeat-visit measurement reproducibility was not assessed.
Single-site, cross-sectional sample of healthy volunteers with severe underrepresentation of older adults (only 16 participants aged ≥70 years, where MAE rose to 15.3-16.0 years).
Sex was excluded from the primary waveform-only model despite modestly improving cross-validated performance when added.
Construct validity: the RF output is a Doppler-based chronological-age estimate, not a validated vascular-age biomarker, as no pulse-wave velocity, intima-media thickness, or clinical outcome data were available.
Non-uniform age distribution and prediction error: 41% of participants (92/227) were in their twenties, and only 16 participants were aged ≥70 years, resulting in higher MAE (15.3–16.0 years) and underpowered performance in older decades.
Incomplete covariate data: anthropometric and body-composition measurements were missing for a large subset, and participants with complete body-composition data (n=88) were significantly younger than those without (31.4 vs. 45.7 years, p = 7.6 × 10⁻¹¹).
Lack of independent physiological reference measurements to confirm the direct physiological interpretation of composite indices (RI, VRI, VEI) or head-to-head benchmarking against other wearable modalities like PPG.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

VRI dominance in Doppler age models extends waveform biomarker research; leaves open validation before any clinical risk use.

Abstract

Background Physiological sensing and machine learning are increasingly combined for health assessment, but many systems emphasize predictive accuracy over physiologically grounded explanation. Objective To describe a small explainable biomarker-discovery framework linking signal processing, explainable machine learning, and physiological interpretation, illustrated and stress-tested with carotid Doppler waveforms. Methods In 227 healthy adults (20–89 years) with portable carotid Doppler recordings, eight waveform features were related to chronological age using Pearson correlation and Random Forest (RF) regression with repeated 10-fold cross-validation. Feature importance was assessed by Gini and permutation methods (in-sample and held-out), and ablation quantified the unique contribution of the top-ranked feature, the Velocity Reflection Index (VRI). Results VRI correlated most strongly with age (r = 0.739, p < 0.001). The RF model estimated age with MAE 8.27 ± 1.40 years and R² = 0.609 ± 0.140. VRI dominated both Gini importance (50.4%) and held-out permutation importance (an even larger share), yet removing VRI and retraining cost only 0.043 R² (0.609 to 0.566), since VRI correlates strongly with other retained landmarks (r = -0.64). A missingness-bias check showed complete-body-composition participants were significantly younger than others (31.4 vs. 45.7 years, p < 0.001). Conclusion Carotid Doppler waveforms carry substantial, explainably identifiable, age-related information, but a single importance metric can overstate a correlated feature's unique contribution; combining importance methods with ablation gives a more defensible account.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Azran Azhim (2026) conducted a cross-sectional in Chronological aging / Vascular aging in healthy adults (n=227). Carotid Doppler waveform biomarkers (8 features including Velocity Reflection Index [VRI]) vs. 7-feature subset without Velocity Reflection Index (VRI) in ablation analysis was evaluated on Chronological age estimation accuracy and correlation of Velocity Reflection Index (VRI) with age (r = 0.739 (VRI correlation with age); RF model R² = 0.609 ± 0.140, MAE = 8.27 ± 1.40 years, p=<0.001). Carotid Doppler waveform features estimated chronological age with R² = 0.609 and MAE 8.27 years, while Velocity Reflection Index correlated strongly with age.

synapsesocial.com/papers/6ae92596a8a6b2adf62f6052 https://doi.org/10.21203/rs.3.rs-11505581/v1

View Full Paper

Ask AI

Bookmark

Share

Explainable Machine-Learning Analysis of Carotid Doppler Waveform Biomarkers for Chronological-Age Estimation

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Explainable Machine-Learning Analysis of Carotid Doppler Waveform Biomarkers for Chronological-Age Estimation

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading