Robust peak recognition in intracranial pressure signals

Area of Science:

• Neuroscience
• Biomedical Engineering
• Medical Signal Processing

Background:

• Intracranial pressure (ICP) pulse waveform morphology is crucial for monitoring and forecasting intracranial and cerebrovascular conditions.
• Existing ICP pulse analysis methods struggle with abnormal or challenging pulse morphologies, leading to performance degradation.

Purpose of the Study:

• To introduce MOCAIP++, a generalized ICP pulse processing framework.
• To investigate the impact of including automatically identified challenging pulses in training peak recognition models.

Main Methods:

• Developed MOCAIP++, integrating multiple peak recognition methods and ICP features.
• Incorporated challenging ICP pulses into the training datasets for peak recognition models.

Main Results:

• Experiments on large ICP datasets and challenging pulses demonstrated significant improvements in peak recognition.
• Both framework generalization and the inclusion of challenging pulses proved complementary and effective.

Conclusions:

• The MOCAIP++ framework enables more reliable extraction of ICP waveform statistics from challenging pulses.
• This advancement facilitates better investigation into the predictive power of ICP pulses for patient outcomes.