Key result
The FFDM-driven Graph Attention and Transformer framework achieved mean absolute errors of 2.12 mmHg for systolic BP and 1.47 mmHg for diastolic BP, satisfying AAMI and BHS standards.
Why the study?
Continuous cuff-less BP monitoring from PPG is challenged by motion artifacts, non-stationarity, and limited physiological interpretability of learned models.
Does the FFDM-GAT-transformer framework accurately estimate cuffless blood pressure from PPG signals?
Does the FFDM-GAT-transformer framework accurately estimate cuffless blood pressure from PPG signals?
Effect estimate: MAE 2.12 mmHg (systolic) and 1.47 mmHg (diastolic)
The FFDM-GAT-transformer framework provides accurate and interpretable cuffless blood pressure estimation from PPG signals, meeting established clinical standards.
May support cuffless BP monitoring research; leaves open prospective clinical validation before any practice consideration.
Cardiovascular diseases (CVDs) remain one of the leading causes of mortality worldwide, with blood pressure (BP) serving as a critical marker for early diagnosis and prevention. Continuous cuff-less BP monitoring from photoplethysmography (PPG) is challenged by motion artifacts, non-stationarity, and limited physiological interpretability of learned models. This work proposes a Fast Fourier Decomposition Method (FFDM)–driven Graph Attention and Transformer framework for robust BP estimation from PPG signals. FFDM is employed to decompose corrupted PPG into Fourier Intrinsic Band Functions, and physiologically meaningful features are extracted from the dominant intrinsic component to form graph-structured representations. A Graph Attention Network models inter-segment dependencies, while a Transformer encoder captures long-range temporal relationships. Model interpretability is investigated using SHapley Additive exPlanations (SHAP). The proposed method is evaluated using subject-wise cross-validation on multiple clinical datasets and achieves mean absolute errors of 2.12 mmHg for systolic BP and 1.47 mmHg for diastolic BP, while satisfying AAMI and BHS standards. These results demonstrate the effectiveness of FFDM-based decomposition and attention-driven learning for accurate and interpretable cuff-less BP estimation.
No takes yet. Share an insight, caveat, or question.
Pankaj et al. (2026) studied Blood pressure estimation. FFDM-driven Graph Attention and Transformer framework was evaluated on Mean absolute error for systolic and diastolic blood pressure (MAE 2.12 mmHg (systolic) and 1.47 mmHg (diastolic)). The FFDM-driven Graph Attention and Transformer framework achieved mean absolute errors of 2.12 mmHg for systolic BP and 1.47 mmHg for diastolic BP, satisfying AAMI and BHS standards.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: