A Hardware-based Lightweight ANN for Real-time Wearable Blood Pressure Estimation

Area of Science:

• Biomedical Engineering
• Cardiovascular Health Technology
• Artificial Intelligence in Healthcare

Background:

• Blood pressure (BP) monitoring is crucial for cardiovascular health and hypertension prevention.
• There is a growing demand for low-power wearable devices for continuous BP estimation.
• Existing methods often face challenges in power consumption and miniaturization for wearable applications.

Purpose of the Study:

• To propose and implement a lightweight artificial neural network (ANN) for accurate and low-power wearable blood pressure (BP) estimation.
• To integrate the ANN model onto an ultra-low-power application-specific integrated circuit (ASIC).
• To evaluate the performance and power efficiency of the developed ASIC for BP monitoring.

Main Methods:

• Development of a lightweight artificial neural network (ANN) model tailored for BP estimation.
• Implementation of the ANN model on an ultra-low-power application-specific integrated circuit (ASIC) using a 0.18 µm CMOS process.
• Validation of the ASIC's performance using standard metrics like Mean Absolute Error (MAE), Standard Deviation (SD), and R-squared (R²) for systolic and diastolic BP estimation.

Main Results:

• The ASIC achieved a low power consumption of 19.72 µW and occupied an area of 730 µm x 730 µm.
• On the test set, the estimated systolic and diastolic BP showed MAE ± SD of 2.47 ± 3.48 mmHg and 1.45 ± 1.88 mmHg, respectively.
• With 8-bit quantization, the MAE ± SD for systolic and diastolic BP were 12.41 ± 5.32 mmHg and 6.29 ± 3.03 mmHg, with an overall R² of 0.9702.

Conclusions:

• The proposed lightweight ANN implemented on an ultra-low-power ASIC offers a viable solution for efficient wearable blood pressure estimation.
• The developed ASIC demonstrates high accuracy and low power consumption, suitable for continuous monitoring applications.
• This technology contributes to advancements in non-invasive hypertension management and personalized cardiovascular health tracking.