Indirect Cardiac Sensing for Noninvasive Pressure Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical devices face challenges in accurately estimating left ventricular pressure without invasive sensing, which is crucial for monitoring cardiac health and delivering therapies like CRT and ICD, due to the sensitivity and confined nature of the human heart.
Innovation Solution
Implementing machine learning techniques, such as decision tree-based models, in medical devices like pacemakers to estimate left ventricular pressure non-invasively using mechanosensory signals, eliminating the need for direct sensing and reducing resource burden.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If invasive sensors are positioned within the left side of the heart to directly sense pressure, then measurement precision of left ventricular pressure is improved, but device complexity and invasiveness increase
Solution Approach 1:
The patent uses an intermediary machine learning model that translates easily acquired mechanosensory signals (accelerometer data, EGMs) into estimates of left ventricular pressure. This intermediary computational layer allows the system to obtain accurate pressure estimates without requiring direct invasive pressure sensing, thereby resolving the contradiction between measurement precision and device complexity.
Solution Approach 2:
The patent replaces the mechanical invasive pressure sensing system with a computational approach using machine learning models. Instead of physically positioning pressure sensors within the heart chambers, the system uses processed mechanosensory signals and EGMs fed into trained models to estimate pressure, substituting mechanical intrusion with computational analysis.
2Measurement precision
If machine learning models are implemented in implantable medical devices, then estimation accuracy of physiological parameters is improved, but resource burden on the device increases
Solution Approach 1:
The patent applies preliminary action by training the machine learning models externally before deployment. The models are trained offline using comprehensive datasets, and only the trained model parameters are implanted in the device. This preliminary training phase allows the device to use lightweight inference algorithms that consume minimal energy during operation, resolving the contradiction between estimation accuracy and energy consumption.
Solution Approach 2:
The patent extracts the computationally intensive training process from the implantable device and performs it externally. Only the essential model parameters and inference logic are retained in the device, separating the heavy computational burden from the resource-constrained implantable system. This extraction allows accurate parameter estimation while maintaining low energy consumption in the actual device.
3Reliability
If direct pressure sensing is used to monitor cardiac function, then reliability of cardiac monitoring is improved, but ease of operation and patient comfort deteriorate
Solution Approach 1:
The patent introduces an intermediary computational system that processes easily obtained mechanosensory signals and EGMs to derive reliable cardiac function metrics. This intermediary approach maintains monitoring reliability by using validated machine learning models while avoiding the operational complexities and patient discomfort associated with invasive pressure sensor positioning and maintenance.
Data Source
AI summary
This disclosure is directed to systems and techniques for detecting change in patient health based upon patient data. In some examples, a medical system includes a mechanosensor configured to sense a first physiological parameter signal of a patient; and processing circuitry configured to: determine one or more features of the first physiological parameter signal; apply a machine leaning model to the one or more features of the first physiological parameter signal; and based on the application of the machine learning model, determine an estimated value of a second physiological parameter.