PaperlayerView exact revision on arXiv

Source-linked AI summary

A Multidimensional Data-Driven Hybrid Transformer Framework for Non-invasive Continuous Blood Pressure Prediction

Yuexin Ma, Jingqi Hou, Yuxuan Kang, Zhaoying Liu

arXiv:2608.23276v1cs.LG

TL;DR

The paper addresses cuffless continuous BP estimation by replacing intermittent, burdensome cuff measurements with a feature-sequence model using ECG/PPG descriptors and demographics. It combines Transformer, KAN, and XGBoost branches with dynamic fusion and reports strong agreement on a segment-level MIMIC-III test split, while retaining important validation boundaries.

  • Problem

    Cuff-based BP measurement is intermittent and limited for long-term continuous monitoring because of discomfort, bulkiness, and operational constraints.

  • Method

    The framework models standardized 10-step sequences of six ECG/PPG-derived descriptors and two demographic covariates using hybrid MSEM and DCFD modules with Transformer, KAN, and XGBoost branches.

  • Results

    The model achieved the best overall agreement-oriented performance among compared baselines and met numerical AAMI mean-error and SD criteria and BHS Grade A percentage thresholds on the segment-level test set.

  • Takeaways & Limitations

    The findings support the framework's agreement with reference BP on the evaluated MIMIC-III split, with component ablation and visualization providing additional evaluated-data evidence.

  • Takeaways & Limitations

    Because segments from the same subject could appear in different partitions, the segment-level split may produce more optimistic generalization estimates than subject-disjoint evaluation.

Abstract

from arXiv · show

Objective. To develop and evaluate a cuffless continuous blood pressure (BP) estimator using temporal physiological and demographic features. We propose a hybrid Transformer framework to estimate diastolic and systolic BP from ECG/PPG-derived feature sequences. Approach. Rather than raw waveforms, the framework models 10-step sequences of six physiological descriptors and two demographic covariates. A Multi-Source Temporal Encoder Module combines Transformer, Kolmogorov-Arnold Network, and XGBoost branches to capture complementary temporal, nonlinear, and tabular information. A Dynamic Conditional Fusion-Decoder applies differential multi-head attention, token-weighted aggregation, and gated residual correction. A robust composite objective jointly optimizes DBP and SBP. Main results. Using the MIMIC-III Waveform and Clinical Databases, the source pool comprised 28,486 waveform segments from 203 subjects, and feature generation retained 53,621 observations from 166 subjects. On 2,431 segment-level held-out test windows, mean error +/- standard deviation was 0.41 +/- 3.74 mmHg for diastolic BP and -1.60 +/- 5.95 mmHg for systolic BP, with 95% limits of agreement of [-6.93, 7.74] and [-13.25, 10.06] mmHg, respectively. The proportions within 10 mmHg were 98.48% and 94.36%. The framework achieved the lowest standard deviations and narrowest limits of agreement among the locally retrained baselines. Significance. The feature-sequence fusion framework improved agreement with reference BP and fell within numerical AAMI and BHS Grade A thresholds on this split. This retrospective analysis is not formal device validation; subject-disjoint and external evaluation remain necessary before clinical use.

1 Introduction

Cuff-based BP measurement is intermittent and burdensome for long-term monitoring, motivating cuffless approaches based on physiological signals and engineered descriptors. The proposed framework fuses multivariate feature sequences through complementary temporal, nonlinear, and tabular branches with dynamic cross-branch decoding and robust joint optimization.

  • Cuff-based BP methods provide intermittent measurements and are constrained by discomfort, bulkiness, and operational demands for long-term monitoring.
  • PTT links ECG cardiac activation to peripheral PPG pulse arrival and is widely used as a BP surrogate, while auxiliary descriptors can improve estimation precision.
  • The proposed pipeline fuses ECG/PPG-derived physiological descriptors with demographic features instead of feeding raw waveform samples directly into the model.
  • MSEM integrates Transformer, KAN, and XGBoost branches to capture complementary temporal, nonlinear, and tabular representations of hemodynamic variation.
  • DCFD unifies the branches using differential multi-head attention to reduce redundancy and improve cross-branch fusion efficiency.
  • The composite robust regression objective jointly optimizes DBP and SBP using weighted error, focal-style, quantile, Huber, regularization, and auxiliary-supervision terms.

2 Related Work

Cuffless BP research has progressed from engineered PTT-based features toward multimodal and end-to-end neural models, but subject variability, calibration, deployment cost, and protocol realism remain barriers. The paper addresses these gaps with hybrid KAN and differential-attention fusion.

  • PTT-only models are sensitive to individual differences and vascular compliance, motivating augmentation with PPG morphology, intensity descriptors, and waveform-based learning.
  • CNNs and RNNs: CNNs learn local waveform patterns, while RNNs model historical dependencies but rely on serial computation that can limit very long-range temporal modeling.
  • Transformers and Generative Models: Transformer-based approaches use self-attention for cuffless BP estimation and waveform synthesis, with recent models combining attention with multiscale, Fourier, or graph-based representations.
  • Multimodal Data Fusion: ECG/PPG fusion is regarded as a strategy for improving robustness against motion artifacts and sensor-specific noise.
  • Feature-, Decision-, and Model-level Fusion: Feature-level fusion can overlook nonlinear cross-modal interactions, whereas decision-level fusion aggregates outputs from independent models.
  • Gap Analysis: Real-world deployment remains constrained by protocol realism, calibration drift, population diversity, fairness reporting, subject variability, calibration needs, and wearable resource costs.
  • Proposed Direction: The proposed hybrid framework combines KAN with differential-attention fusion to enhance integration of multibranch features.

3 Methodology

The methodology maps standardized 10-step windows of eight physiological and demographic variables to subsequent DBP and SBP references. Its pipeline combines feature-sequence construction, a Transformer-based multibranch encoder, and a conditional fusion decoder with complementary nonlinear and tabular branches.

  • Problem Definition: Each input contains L = 10 consecutive feature records and F = 8 variables, while the target contains subsequent-record DBP and SBP references.
  • Overall Framework: The framework integrates waveform and metadata construction, physiological descriptor derivation, feature-sequence construction, and hybrid Transformer prediction with composite optimization.
  • Overall Framework: MSEM and DCFD together model temporal dependencies, nonlinear hemodynamic characteristics, and cross-branch feature interactions.
  • Encoding and Branches: The encoding stage removes missing observations, standardizes eight inputs using training-partition statistics, and normalizes each window temporally for neural branches.
  • Encoding and Branches: The Transformer main head uses mean-pooled temporal representations, KAN models nonlinear relationships from the flattened window, and XGBoost supplies a tabular estimate from window statistics.

DCFD MSEM

The framework combines physiological feature-sequence processing with hybrid temporal, nonlinear, and tabular branches, then fuses their representations for continuous BP prediction. Its pipeline derives eight predictors from synchronized signals and metadata, trains joint DBP/SBP regression with a composite robust objective, and applies data-quality and cohort-accounting procedures.

  • MSEM: The MSEM combines Transformer, KAN, and XGBoost branches to model complementary temporal, nonlinear, and tabular representations.Branch projections create common tokens for downstream fusion.
  • DCFD: The DCFD applies differential attention, fused representation learning, token-weighted aggregation, and gated residual correction to integrate branch outputs.The neural correction is added to the two-dimensional XGBoost baseline estimate.
  • Preprocessing and feature engineering: The upstream pipeline uses filtering, robust normalization, polarity correction, and fiducial detection to derive timing and morphology features from synchronized ECG and PPG.PTT, the PPG foot-to-peak interval, RR interval, and HR are defined from ECG R-peaks and PPG landmarks.
  • Model-ready inputs: The model input is a standardized 10-step sequence containing six physiological descriptors plus age and body weight, rather than raw waveform samples.The physiological variables include PTT, foot-to-peak interval, heart rate, maximum PPG slope, RI, and SD time.
  • Data preparation and limitations: Complete-case filtering removed 9 of 53,630 rows with missing weight, while the analytical table retained physiologically unusual values without retrospective exclusion.The retained table included two rows with SBP ≤ DBP and extreme reflection-index values.
  • Composite robust loss: The composite objective jointly optimizes DBP and SBP using target-weighted squared errors, scheduled focal-style and quantile penalties, early Huber stabilization, and regularization.Learnable target-specific scales adjust the relative DBP and SBP squared-error contributions; the Huber term is active for the first ten epochs.

4 Experiments

Experiments compared the hybrid framework with locally retrained baselines and component-removed variants using a shared, segment-level split. The proposed model achieved strongest agreement-oriented performance, while ablations identified XGBoost, KAN, and MHDA as important components, especially for SBP stability.

  • Experimental protocol: The shared 53,621-observation table used an 80%/10%/10% segment split, but subject identifiers were unavailable, so the test set was not subject-disjoint.Windows from different segments of the same subject could therefore occur across partitions.
  • Baseline comparison: 3.74 mmHg was the lowest DBP SD, with the narrowest 95% LoA of [-6.93, 7.74] mmHg among compared methods.DBP accuracies were 87.95%, 98.48%, and 99.01% at ≤5, ≤10, and ≤15 mmHg.
  • Baseline comparison: 5.95 mmHg was the lowest SBP SD, while the proposed model achieved the narrowest SBP 95% LoA of [-13.25, 10.06] mmHg and highest threshold accuracies.SBP accuracies were 70.75%, 94.36%, and 98.11% at the three thresholds, although XGBoost had the bias closest to zero at -0.60 mmHg.
  • Baseline comparison: Compared with XGBoost, the hybrid reduced SBP SD from 7.10 to 5.95 mmHg and improved ≤10 mmHg accuracy from 87.23% to 94.36%.Its SBP LoA also narrowed from [-14.52, 13.32] to [-13.25, 10.06] mmHg.
  • Ablation study: Removing XGBoost increased DBP SD from 3.74 to 12.00 mmHg and SBP SD from 5.95 to 23.43 mmHg, producing much wider LoA intervals.The ablation supports XGBoost as a statistical anchor in the hybrid residual framework.
  • Ablation study: Removing KAN reduced SBP ≤10 mmHg accuracy from 94.36% to 40.07%, while removing MHDA reduced it to 42.62%.The MHDA ablation also widened SBP LoA to [-17.37, 37.27] mmHg, whereas the KAN ablation widened it to [-22.43, 36.95] mmHg.
  • Ablation study: The full model had the narrowest DBP and SBP LoA intervals among reduced variants and the best SBP cumulative accuracy at all three thresholds.Its DBP ≤5 and ≤10 mmHg accuracies were also best, despite a slightly higher DBP ≤15 mmHg accuracy for one reduced variant.
  • Agreement analysis: Bland–Altman analysis showed biases of 0.41 mmHg for DBP and −1.60 mmHg for SBP, with unadjusted 95% LoA of [-6.93, 7.74] and [-13.25, 10.06] mmHg.These are descriptive per-window estimates for 2,431 test windows, which are not statistically independent because of repeated subjects and adjacent windows.

5 Discussion

The hybrid framework showed strong agreement-oriented performance on the evaluated segment-level test set, but several design and evaluation limitations constrain claims about clinical or subject-independent generalization.

  • The proposed framework achieved the best overall agreement-oriented performance among the compared continuous BP estimation models.It explicitly combines six ECG/PPG-derived descriptors with age and body weight, unlike end-to-end waveform models that rely solely on latent feature extraction.
  • The model met numerical AAMI mean-error and standard-deviation criteria and numerical BHS Grade A percentage thresholds on the held-out test set.These comparisons motivate formal clinical validation but do not constitute compliance testing, device validation, or certification.
  • MIMIC-III ICU data may not represent ambulatory conditions, where motion artifacts can reduce signal quality and robustness requires further validation.The authors also note that periodic calibration could further improve long-term accuracy and propose validation on real-world wearable datasets.
  • The segment-level split allowed segments from the same subject to appear in different partitions, potentially producing more optimistic generalization estimates than subject-disjoint evaluation.Subject mapping was reconstructed later for cohort accounting but was not used to alter the established partitions.
  • Future evaluation should use a de novo subject-disjoint split defined before scaling and window construction, together with external-cohort validation.
  • The complete-case pipeline lacked prespecified physiological-range exclusions or RI clipping, and the influence of retained extreme values was not quantified through retrained sensitivity analysis.Future work should compare results with and without physiologically justified quality-control exclusions.
  • Reported SD and Bland–Altman limits of agreement are descriptive per-window estimates because temporally adjacent and repeated observations were not analyzed with subject-clustered inference.The 2,431 test windows therefore should not be interpreted as subject-level precision estimates.

6 Conclusion

The paper presents a hybrid, multidimensional feature-sequence framework for cuffless continuous BP monitoring. On MIMIC-III, it showed strong agreement-oriented performance, but broader clinical generalization requires more rigorous evaluation.

  • The framework models standardized 10-step sequences of six ECG/PPG-derived descriptors and two demographic covariates using hybrid Transformer components with complementary XGBoost and KAN representations.
  • On MIMIC-III, the model achieved the best overall agreement-oriented performance among the compared baselines.The reported numerical AAMI and BHS comparisons were obtained on a segment-level test set.
  • The numerical AAMI and BHS comparisons are descriptive and do not constitute formal compliance testing or device certification.The conclusion calls for prespecified quality-control rules, repeated-measures inference, de novo subject-disjoint evaluation, and external-cohort validation.

Ethical statement

This study used de-identified secondary data from MIMIC-III databases rather than recruiting participants or collecting new physiological or clinical data.

  • The analysis used de-identified physiological and linked demographic data from the MIMIC-III Waveform and Clinical Databases.
  • No participants were recruited and no new physiological or clinical data were collected by the authors.

Conflict of interest

The paper reports author roles and states that all authors reviewed and approved the final manuscript.

  • The authors declare no competing interests.
  • Author contributions covered conceptualization, methodology, software, data curation, validation, analysis, visualization, supervision, resources, and manuscript preparation.
  • All authors reviewed and approved the final manuscript.
Loading 2608.23276v1…