AI/ML Classification for Spinal Cord Stimulation Signal Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neuromodulation therapies, particularly those using closed-loop systems, face challenges in efficiently identifying and classifying evoked compound action potentials (ECAPs) from recorded signals, leading to time-consuming setup processes and varying success rates.
Innovation Solution
The system employs artificial intelligence (AI) and machine learning (ML) methods to create classification models that reduce the complexity of identifying ECAPs in recorded signals. These models support the determination of signal classifications, such as ECAP, non-ECAP, and noise, to inform therapy optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional signal classification methods are used to identify ECAPs from recorded signals, then the system can detect neural responses, but the process becomes time-consuming and has varying success rates
Solution Approach 1:
The patent transforms the ECAP detection problem by changing the parameter space from raw time-domain signals to frequency-domain representations (spectrograms). This transformation enables machine learning models to automatically identify ECAP characteristics, achieving over 90% accuracy while reducing setup time through automated classification
Solution Approach 2:
The patent replaces manual signal analysis methods with automated machine learning-based classification systems. The ML models process recorded signals to automatically distinguish ECAPs from artifacts and noise, eliminating time-consuming manual verification and achieving consistent high success rates
2Reliability
If manual signal analysis is used to classify recorded signals, then clinicians can identify ECAPs, but the setup process becomes complex and time-consuming
Solution Approach 1:
The patent implements self-service through automated machine learning models that independently classify recorded signals without requiring manual intervention. The system automatically processes raw signals, extracts features, and identifies ECAPs with high reliability, eliminating the need for complex manual signal processing procedures
Solution Approach 2:
The patent introduces machine learning models as intermediaries between the recorded signals and the classification decision. These models act as intelligent mediators that automatically distinguish ECAPs from artifacts and noise, simplifying the overall system while improving reliability through consistent automated classification
3Productivity
If automated classification is implemented to reduce setup time, then efficiency improves, but accuracy in distinguishing ECAPs from artifacts must be maintained
Solution Approach 1:
The patent applies preliminary action by preprocessing recorded signals into frequency-domain representations (spectrograms) before classification. This preparation step enhances the distinguishability of ECAP features, enabling automated models to achieve high accuracy quickly without requiring complex real-time analysis
Solution Approach 2:
The patent incorporates feedback mechanisms where the automated classification system continuously learns from classified results to improve future classifications. This feedback loop maintains and enhances accuracy over time while preserving the efficiency gains from automation
Data Source
AI summary
A system, device, and method support receiving a data signal from one or more sensors associated with the system in response to therapy delivered to a patient. The system, device, and method support assigning a classification to one or more portions of a waveform associated with the data signal based on characteristic information associated with the one or more portions of the waveform. The system, device, and method support providing, based on the classification, one or more parameters associated with delivering the therapy.