A Delineator for Arterial Blood Pressure Waveform Analysis Based on a Deep Learning Technique

Area of Science:

• Biomedical engineering
• Artificial intelligence in healthcare
• Signal processing

Background:

• Accurate blood pressure monitoring is crucial for cardiovascular health.
• Existing methods for detecting blood pressure signals can be limited by artifacts and noise.
• Automated detection of valid blood pressure points is needed for reliable analysis.

Purpose of the Study:

• To implement a deep learning technique for blood pressure detection point identification.
• To evaluate the performance of semantic segmentation models against a reference detector.
• To assess the models' ability to differentiate valid signals from artifacts.

Main Methods:

• Developed and trained two deep learning models based on semantic segmentation.
• Evaluated model performance on classic benchmarks and a modified public database.
• Tested models on signals with realistic test maneuvers and artifacts.

Main Results:

• The proposed deep learning methodology outperformed the reference detector on two of three classic benchmarks.
• Models demonstrated superior performance on a public database with introduced artifacts.
• Both models successfully differentiated regions containing valid information from artifact regions.

Conclusions:

• Deep learning semantic segmentation offers a robust approach for blood pressure detection point delineation.
• The developed models exhibit a unique capacity to identify and exclude artifact regions.
• This technique holds promise for improving the accuracy and reliability of automated blood pressure analysis.