Detection of a Stroke Volume Decrease by Machine-Learning Algorithms Based on Thoracic Bioimpedance in Experimental...

Area of Science:

• Cardiovascular Physiology
• Medical Technology
• Data Science in Medicine

Background:

• Compensated shock and hypovolaemia are often undetected, leading to patient deterioration.
• Current detection methods lack accuracy and non-invasiveness for early identification.
• Automated, non-invasive monitoring is crucial for perioperative and critically ill patients.

Purpose of the Study:

• To develop a predictive model for stroke volume index (SVI) decrease using electrical cardiometry (EC).
• To assess the model's efficacy in detecting simulated central hypovolaemia.
• To compare the predictive power of EC variables against traditional vital signs.

Main Methods:

• Experimental study involving 30 healthy male volunteers.
• Central hypovolaemia simulated using a lower body negative pressure (LBNP) chamber.
• Stroke volume index (SVI) assessed via transthoracic echo (SVI-TTE) as reference.
• Machine-learning algorithm developed using EC variables.
• Comparison of model performance against heart rate and systolic arterial pressure.

Main Results:

• Simulated hypovolaemia caused significant SVI-TTE decline with stable vital signs.
• The EC-based model demonstrated superior predictive ability for SVI decrease (AUC: 0.91).
• EC model significantly outperformed heart rate (AUC: 0.83) and systolic arterial pressure (AUC: 0.82).

Conclusions:

• EC variables, analyzed by machine learning, can accurately predict relevant SVI decreases.
• This approach offers a potential automated, non-invasive method for indicating hypovolaemia and compensated shock.
• The developed model shows promise for early detection and management of critical hemodynamic conditions.