Report on Current Developments in Non-Invasive Blood Pressure Monitoring and Remote Physiological Measurement
General Direction of the Field
The recent advancements in the field of non-invasive blood pressure (BP) monitoring and remote physiological measurement are notably shifting towards more sophisticated deep learning models that enhance the accuracy, robustness, and generalizability of these measurements. The focus is increasingly on developing models that can effectively capture long-range dependencies and dynamic changes in physiological signals, which are crucial for real-time monitoring and early detection of cardiovascular diseases.
One of the key trends is the integration of invertible neural networks (INNs) for signal reconstruction tasks, which aim to prevent information loss by simultaneously learning forward and inverse mappings. This approach is particularly promising for tasks like photoplethysmography (PPG) to arterial blood pressure (ABP) reconstruction, where preserving high-frequency details is essential for accurate BP measurement.
Another significant development is the use of state space models (SSMs) like Mamba, which are being explored for their ability to efficiently model long-range dependencies in physiological signals. These models are being combined with novel architectures, such as dual-stream SlowFast networks, to capture both local and global temporal features, thereby improving the accuracy of remote physiological measurements.
The field is also witnessing a growing emphasis on domain adaptation and generalization, particularly in scenarios where access to source domain data is limited or restricted due to privacy concerns. Source-free domain adaptation (SFDA) frameworks are being introduced to address these challenges, leveraging techniques like spatio-temporal consistency and frequency-domain alignment to enhance model performance across different domains.
Noteworthy Innovations
- INN-PAR: Introduces an invertible neural network for PPG to ABP reconstruction, significantly outperforming state-of-the-art methods in waveform reconstruction and BP measurement accuracy.
- PhysMamba: Proposes a Mamba-based framework for efficient remote physiological measurement, demonstrating superior performance and efficiency on benchmark datasets.
- SFDA-rPPG: Presents the first source-free domain adaptation benchmark for rPPG measurement, effectively aligning power spectrum distributions across domains without access to source data.
