Artificial Intelligence-Driven Prognosis of Respiratory Mechanics: Forecasting Tissue Hysteresivity Using Long...

Area of Science:

• Respiratory Medicine
• Biomedical Engineering
• Artificial Intelligence

Background:

• Tissue hysteresivity is a crucial indicator for respiratory disease progression, typically measured via forced oscillation techniques.
• Current measurement protocols can be lengthy and demanding for patients.
• Utilizing multivariate data and AI for respiratory monitoring is an emerging area with untapped potential.

Purpose of the Study:

• To reduce the number and duration of measurements for tissue hysteresivity.
• To explore the use of AI, specifically Long Short-Term Memory (LSTM) networks, for estimating and forecasting tissue hysteresivity.
• To integrate data from Forced Oscillation Technique (FOT) devices and the Equivital (EQV) LifeMonitor for enhanced respiratory analysis.

Main Methods:

• Collected synchronized data from EQV LifeMonitor, FOT prototype, and RESMON devices over a two-hour protocol.
• Employed LSTM models to estimate hysteresivity (η) using continuous heart rate (HR) data.
• Utilized LSTM to forecast η by first predicting HR from electrocardiogram (ECG) data.

Main Results:

• LSTM accurately estimated tissue hysteresivity (η) with R2 = 0.851 and MSE = 0.296.
• LSTM successfully forecasted η with R2 = 0.883 and MSE = 0.528.
• The number of required measurements was reduced by a factor of three, from ten to three.

Conclusions:

• AI-driven integration of FOT and wearable sensor data offers a promising method for accurate tissue hysteresivity estimation and forecasting.
• This novel approach significantly reduces patient burden, measurement time, and associated costs.
• The study highlights the potential of AI in advancing ambulatory respiratory monitoring.