Learning Apparatus for Cardiac Sound Time Series Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the condition of a heart from cardiac sound time series have poor accuracy, requiring labor-intensive supervised data creation with electrocardiographs and struggling to decompose time series into linear sums of fluctuating oscillators accurately.
Innovation Solution
A learning apparatus and method that acquire and analyze time series represented by a linear sum of amplitude waveforms of fluctuating oscillators using a linear sum estimation learning model, which updates based on probabilistic state transitions and symbol outputs to improve decomposition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used to estimate heart conditions from cardiac sound time series, then the analysis can be performed, but the accuracy of decomposition into linear sum of fluctuating oscillators is poor
Solution Approach 1:
The patent implements feedback through iterative optimization where the learning model continuously refines its parameters based on the observed time series data. The model adjusts its decomposition parameters to minimize the difference between reconstructed and actual time series, thereby improving measurement precision of the decomposition accuracy and reliability of heart condition estimation.
Solution Approach 2:
The patent changes parameters of the learning model dynamically during the optimization process. The model adapts its decomposition parameters, oscillation frequency parameters, and amplitude parameters based on the input time series characteristics, enabling accurate decomposition even for complex non-stationary cardiac sound signals.
2Quantity of substance
If supervised data creation using electrocardiograph is performed manually, then training data can be obtained, but the process requires significant labor
Solution Approach 1:
The patent enables self-service by allowing the learning model to automatically generate training data through simulation. The model can create synthetic cardiac sound time series with known ground truth labels by simulating the physical processes of heart valve vibrations and blood flow, eliminating the need for manual annotation of electrocardiograph data while providing abundant training examples.
Solution Approach 2:
The patent creates copies of real cardiac sound data through synthetic data generation. By modeling the physical mechanisms of heart sounds and generating virtual time series that replicate real patient data characteristics, the system obtains numerous training samples without requiring actual patient recordings or manual processing of electrocardiograph data.
3Measurement precision
If the time series is decomposed into linear sum of fluctuating oscillators, then the analysis can be performed, but the decomposition accuracy is insufficient for reliable heart condition estimation
Solution Approach 1:
The patent achieves universality by developing a single learning model that can accurately decompose various types of cardiac sound time series with different pathologies and physiological states. The model handles multiple oscillation modes, frequency variations, and amplitude changes within a unified framework, improving decomposition accuracy without requiring separate specialized algorithms for different cardiac conditions.
Data Source
AI summary
An aspect of the present invention is a learning apparatus including a time series acquisition unit that is configured to, with a time series of an amplitude of a fluctuating oscillator whose amplitude changes periodically being defined as an oscillator time series, acquire an observed time series which is a time series represented by an oscillator linear sum which is a linear sum of the oscillator time series and a learning processing execution unit that is configured to use an expression representing a generation mechanism of the observed time series and a mathematical model representing a relationship between a probabilistic state transition of a state of a generation source of the observed time series and a symbol output which is information probabilistically output in the state to execute a linear sum estimation learning model which is a mathematical model that is configured to estimate the oscillator linear sum of the observed time series on the basis of the observed time series, wherein the learning processing execution unit is configured to update the linear sum estimation learning model on the basis of a result of execution of the linear sum estimation learning model.