Streaming Wavelet Transform for Respiratory Waveform Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing implantable cardiac devices face challenges in accurately detecting respiratory waveforms due to sources of error like aliasing and non-stationary physiological waveforms, which conventional digital filtering methods fail to address effectively, leading to energy consumption and complex circuit designs.
Innovation Solution
An implantable cardiac device that performs streaming Wavelet transformation on intrathoracic or intracardiac impedance, pressure, or accelerometry input streams to separate respiratory waveforms, using an adaptable sampling frequency and a filter bank that allows for real-time source separation without matrix processing, preserving fiducial points and reducing energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional digital filtering is used to extract respiratory waveforms, then the filtering process can be implemented, but it distorts the physiological waveform and fails to adapt to non-stationary respiration characteristics
Solution Approach 1:
The patent applies wavelet transform with variable decomposition levels and adaptive thresholding parameters that change based on the non-stationary characteristics of respiratory signals. The wavelet coefficients are computed using scaling functions and wavelet functions with adjustable parameters, allowing the filtering process to adapt to changing signal characteristics without distorting the physiological waveform.
Solution Approach 2:
The invention uses dynamic adaptive filtering where the filter characteristics change over time based on the input signal properties. The wavelet transform provides time-frequency localization that adapts to the non-stationary nature of respiration, allowing the system to track and extract respiratory waveforms accurately without imposing fixed filter constraints that would distort the signal.
2Adaptability or versatility
If multiple adaptive filter designs are implemented to handle changing waveform components, then adaptability to different respiration states is improved, but energy consumption increases and circuit design becomes complicated
Solution Approach 1:
The patent segments the respiratory signal extraction process into distinct wavelet decomposition stages, where each stage processes specific frequency bands independently. This segmentation allows the system to focus computational resources only on relevant signal components rather than processing the entire spectrum, reducing overall energy consumption while maintaining adaptability to different respiration states.
Solution Approach 2:
The invention extracts only the necessary respiratory waveform information from the composite signal using wavelet coefficient analysis. By taking out and processing only the relevant frequency components that contain respiratory information, the system avoids the energy-intensive operation of processing all signal components, thereby reducing power consumption while maintaining extraction accuracy.
3Ease of operation
If adaptive filtering is used with assumption of stationarity, then the filtering can be performed, but convergence fails during adaptation due to non-stationary physiological waveforms
Solution Approach 1:
The patent employs dynamic wavelet-based filtering that continuously adapts to the non-stationary characteristics of physiological signals. The wavelet transform provides time-varying frequency analysis that tracks signal changes, ensuring reliable convergence without requiring the unrealistic assumption of stationarity. The filter dynamically adjusts its parameters based on the instantaneous signal characteristics.
4Measurement precision
If wavelet transform is implemented for source separation of respiration, then accurate extraction of respiratory waveforms is achieved, but implementation cost in implantable devices becomes very high
Solution Approach 1:
The patent segments the wavelet transform implementation into a filter bank structure with separate processing channels for different decomposition levels. This segmentation allows parallel processing of independent frequency bands, reducing computational complexity compared to a monolithic wavelet transform implementation, while maintaining the accuracy benefits of wavelet-based source separation.
Solution Approach 2:
The invention applies partial wavelet decomposition by selecting only the necessary decomposition levels and frequency bands relevant to respiratory signal extraction. Rather than performing complete multi-level decomposition of all signal components, the system applies wavelet transform only to the extent needed for respiratory waveform separation, reducing implementation complexity while preserving measurement precision.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to the present invention, an implantable cardiac device 80 is provided. The implantable cardiac device 80 comprises a sensing unit 70 adapted to measure an intrathorac-ic or intracardiac impedance, pressure, and/or accelerometry input stream 69, which comprises a patient's respiratory waveforms. Furthermore, the implantable cardiac device 80 comprises a quantizer-unit 60 adapted to sample the input stream 69 with an initial sampling frequency Fs, providing input samples 65 of the input stream. The implantable cardiac device 80 further comprises a filter bank 50 suited to perform a streaming Wavelet transformation on the input samples 65 on a sample-by-sample basis, using the initial sampling frequency Fs provided by the quantizer-unit 60, wherein the streaming Wavelet transformation is adapted to perform a source separation, extracting, and separating the respiratory waveforms of the input stream 69.