A Novel Automated Blood Pressure Estimation Algorithm Using Sequences of Korotkoff Sounds

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Signal Processing

Background:

• Automated non-invasive blood pressure (NIBP) monitoring is increasingly used for home BP measurement.
• Current automated NIBP devices predominantly use the oscillometric technique, with fewer employing the auscultatory technique.
• Deep learning (DL) shows promise in data classification and feature extraction for medical applications.

Purpose of the Study:

• To propose a novel automated artificial intelligence (AI)-based technique for NIBP estimation from auscultatory waveforms (AWs).
• To convert NIBP estimation into a sequence-to-sequence classification problem solvable with DL.
• To evaluate the performance of the proposed AI algorithm against established NIBP measurement standards.

Main Methods:

• Segmenting AWs and applying wavelet packet decomposition to extract features.
• Labeling segments as between systolic and diastolic phases or otherwise.
• Utilizing a bidirectional long short-term memory recurrent neural network (BiLSTM-RNN) for sequence-to-sequence classification.
• Employing a 5-fold cross-validation scheme on a dataset of 350 NIBP recordings.

Main Results:

• The proposed AI-based technique achieved an average mean absolute error of 1.7±3.7 mmHg for systolic BP (SBP) and 3.4 ±5.0 mmHg for diastolic BP (DBP).
• Performance was evaluated relative to reference blood pressure values.
• The DL algorithm demonstrated high accuracy in classifying auscultatory waveform segments.

Conclusions:

• The developed automated BP estimation algorithm, leveraging DL and AWs, offers significant benefits for NIBP monitoring.
• The novel sequence-to-sequence classification approach provides a viable alternative for automated auscultatory BP measurement.
• This AI-driven method has the potential to enhance the accuracy and accessibility of home BP monitoring.