Quantifying the Suitability of Biosignals Acquired During Surgery for Multimodal Analysis

Area of Science:

• Biomedical engineering
• Signal processing
• Intraoperative monitoring

Background:

• Large surgical biosignal datasets are now available, enabling multimodal analysis.
• Multimodal analysis offers deeper insights than unimodal analysis but data suitability is unknown.
• Manual data inspection for suitability is infeasible due to data volume.

Purpose of the Study:

• To evaluate the suitability of intraoperatively acquired biosignals for multimodal analysis.
• To compare the suitability of surgical data (VitalDB) with sleep data (SIESTA).

Main Methods:

• Applied unimodal signal quality indicators to electrocardiography, electroencephalography, and respiratory signals.
• Analyzed data in 10-second segments.
• Considered a segment suitable for multimodal analysis if all three signals passed quality indicators.

Main Results:

• 72% of the VitalDB surgical dataset and 82% of the SIESTA sleep dataset were suitable.
• Unsuitable segments showed constant or physiologically unreasonable signal values.
• Signal quality distributions were similar between datasets, with potential biases from different setups.

Conclusions:

• The majority of VitalDB surgical data is suitable for multimodal analysis.
• This finding supports the use of multimodal analysis in surgical settings.
• Signal quality indicators are effective in assessing data suitability.