WGAN-BP: A Blood Pressure Estimation Method Based on Wasserstein Generative Adversarial Network

Le résumé fourni par la source

Continuous blood pressure monitoring is crucial for patients with hypertension and cardiovascular diseases. To address the issue that traditional methods cannot continuously measure arterial blood pressure, this study proposes a blood pressure estimation model based on Wasserstein Generative Adversarial Network(WGAN-BP), which estimates arterial blood pressure(ABP) from photoplethysmography(PPG) waveforms. The generator of the model employs a UNet structure composed of recurrent residual modules, while the discriminator takes PPG as a conditional input, guiding the network training process. Additionally, this study introduces a time-frequency hybrid loss function as a part of generator's loss function, which takes into account both the temporal and spectral features of the ABP waveforms. The model was trained on the MIMIC-II (UCI) dataset and its performance in estimating blood pressure waveforms outperformed existing studies.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.