Generalizable Blood Pressure Estimation from Multi-Wavelength PPG Using Curriculum-Adversarial Learning

TL;DR

This work tackles generalizable cuff-less BP estimation from multi-wavelength PPG by addressing data leakage through strict subject-level splits and introducing Curriculum-Adversarial Learning that jointly performs hypertension classification and BP regression while learning subject-invariant representations. The framework fuses signals across four wavelengths via attention-based channels and employs domain-adversarial training with a gradient reversal layer, guided by a curriculum that shifts from classification to regression. On a public four-wavelength PPG dataset, it achieves SBP/DBP MAEs of $14.2$ mmHg and $6.4$ mmHg, respectively, and outperforms several baselines; ablations confirm the value of both curriculum and adversarial components. The results support the potential of multi-wavelength PPG combined with curriculum-adversarial strategies for robust, generalizable BP estimation, with implications for hypertension screening and continuous monitoring, though performance at extreme BP and in degraded signal conditions remains a challenge.

Abstract

Accurate and generalizable blood pressure (BP) estimation is vital for the early detection and management of cardiovascular diseases. In this study, we enforce subject-level data splitting on a public multi-wavelength photoplethysmography (PPG) dataset and propose a generalizable BP estimation framework based on curriculum-adversarial learning. Our approach combines curriculum learning, which transitions from hypertension classification to BP regression, with domain-adversarial training that confuses subject identity to encourage the learning of subject-invariant features. Experiments show that multi-channel fusion consistently outperforms single-channel models. On the four-wavelength PPG dataset, our method achieves strong performance under strict subject-level splitting, with mean absolute errors (MAE) of 14.2mmHg for systolic blood pressure (SBP) and 6.4mmHg for diastolic blood pressure (DBP). Additionally, ablation studies validate the effectiveness of both the curriculum and adversarial components. These results highlight the potential of leveraging complementary information in multi-wavelength PPG and curriculum-adversarial strategies for accurate and robust BP estimation.

Figures (4)

• Figure 1: Overview of blood pressure estimation from multi-wavelength PPG
• Figure 2: Blood pressure distribution (a) SBP (b) DBP
• Figure 3: Data preprocessing and multi-channel feature extraction
• Figure 4: Multi-channel fusion and curriculum-adversarial learning