Stroke Volume Measurement via Neural Network Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining cardiac stroke volume using pulse contour analysis are unreliable and impractical, especially in emergency situations, due to moderate correlation with reference methods and high percentage errors, and fail to accurately track changes in cardiac output, particularly in obese patients.
Innovation Solution
The method improves cardiac stroke volume determination by calibrating pulse contour stroke volume using perfusion parameters through fat free and adipose mass, and incorporating a fluid responsiveness parameter to enhance sensitivity and accuracy, allowing for non-invasive and precise tracking of changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional pulse contour analysis is used to determine cardiac stroke volume, then the measurement is non-invasive and can be performed continuously, but the measurement precision is poor with moderate correlation to reference methods and high percentage errors
Solution Approach 1:
The patent introduces an artificial neural network as an intermediary layer between the raw pulse contour data and the stroke volume calculation. The neural network processes the arterial pressure waveform and derives intermediate physiological parameters (such as ejection phase characteristics, dicrotic notch timing, and area under the curve) that are then used to calculate stroke volume. This intermediary processing significantly improves measurement precision while maintaining the non-invasive nature of the original measurement.
Solution Approach 2:
The patent transforms the conventional pulse contour analysis by changing the parameters used for stroke volume determination. Instead of relying on simple geometric approximations of the pressure waveform, the invention uses neural network-derived parameters including normalized ejection phase area, timing ratios, and calibrated physiological parameters. These parameter changes enable the system to achieve accuracy comparable to invasive reference methods while preserving the non-invasive advantage.
2Measurement precision
If pulmonary artery thermodilution is used to measure cardiac output, then the measurement accuracy is high, but the method is highly invasive with threatening complications and time consuming
Solution Approach 1:
The patent creates a computational model that copies and simulates the physiological processes measured by invasive thermodilution methods. The artificial neural network is trained on data from reference methods and learns to replicate their accuracy. This allows the system to produce stroke volume measurements with precision comparable to pulmonary artery thermodilution without requiring actual catheter insertion or cold fluid injection, thereby eliminating the harmful invasive effects.
Solution Approach 2:
The invention replaces the mechanical invasive measurement system (thermodilution catheters, fluid injection apparatus) with a computational system based on artificial intelligence. The neural network processes standard non-invasive blood pressure waveform data and derives stroke volume through mathematical modeling and pattern recognition, substituting complex mechanical measurement devices with intelligent software-based analysis that achieves similar accuracy without physical intrusion into the patient's vasculature.
3Productivity
If pulse contour analysis is performed without individual patient calibration, then the method is simple and rapid, but the measurement precision deteriorates due to lack of personalization
Solution Approach 1:
The patent implements a preliminary calibration phase where the artificial neural network is trained on individual patient data obtained during an initial measurement period. During this phase, the system learns the patient's specific hemodynamic characteristics, arterial compliance, and waveform patterns. Once calibrated, the neural network can rapidly process subsequent measurements with high individualized accuracy. This preliminary action ensures that the rapid ongoing measurements are both fast and precisely tailored to each patient's physiology.
Solution Approach 2:
The system dynamically adapts to individual patients through the neural network's ability to process and learn from real-time data. The calibration process is not static but continuously refines the model based on incoming measurements and changing physiological conditions. This dynamic adaptation allows the system to maintain high measurement precision across varying patient states while preserving rapid measurement capability, as the neural network automatically adjusts to new conditions without requiring manual recalibration.
Data Source
Figure 1~1a
Figure 1b
Figure 2
AI summary
A method and apparatus is provided for determining a stroke volume (SV) of an individual, comprising the steps of: providing a first pulse contour stroke volume based on one or more characteristics of a measured arterial blood pressure waveform or providing a conventionally derived pulse contour stroke volume, determining at least one perfusion parameter descriptive for the perfusion through the fat free mass and the adipose mass of a body of the individual, and/or determining at least one fluid responsiveness parameter function depending on a fluid responsiveness parameter descriptive for a heart-lung interaction of the individual, and adjusting the first pulse contour stroke volume or the conventionally derived pulse contour stroke volume based on at the least one of the perfusion parameter and/or the fluid responsiveness parameter function to provide a second pulse contour stroke volume.