System and method for pain monitoring using a multidimensional analysis of physiological signals
Inventors
Zuckerman-Stark, Galit • Kliger, Mark
Assignees
Publication Number
US-11259708-B2
Publication Date
2022-03-01
Expiration Date
2028-11-13
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
The present invention is for a method and system for pain classification and monitoring optionally in a subject that is an awake, semi-awake or sedated.
Core Innovation
The invention provides a system and method for pain monitoring and classification that utilizes a multidimensional analysis of a plurality of physiological signals from a subject. It addresses pain detection and monitoring across various states of consciousness, including awake, semi-awake, and sedated states. The system extracts a great plurality of features from signals such as PPG, GSR, skin temperature, ECG, respiration, EMG, and EEG/FEMG, combining and processing these features to improve the accuracy of pain classification through machine learning techniques.
The problem being solved stems from the limitations in existing pain monitoring and Depth of Anesthesia (DOA) systems that predominantly focus on unconscious patients under anesthesia and often rely on analysis of single physiological signals or limited parameters. These prior systems do not sufficiently account for the complexity and multidimensional nature of pain across different consciousness states, nor do they leverage a large plurality of physiological features for improved pain detection and classification. The invention addresses this deficiency by implementing an integrated approach that uses multiple physiological signals, advanced signal processing, feature extraction, feature selection, dimensionality reduction, and classification techniques adapted for various pain states and stimuli.
The invention encompasses a method that includes acquiring multiple physiological signals, preprocessing these signals to enhance quality, extracting numerous features to form a great plurality of features vector, performing normalization and dimensionality reduction on this vector, and finally applying classifiers to categorize pain levels into at least two classes, with potential for multi-class classification and scalable pain scoring. The system may further incorporate a priori data such as patient history or environmental factors to enhance classification accuracy, providing a comprehensive and adaptable solution for pain monitoring in clinical environments.
Claims Coverage
The independent claim covers a method for pain monitoring that involves obtaining physiological signals, processing and extracting features, transforming these features, and classifying the pain status using a specific classification algorithm.
Obtaining multiple physiological signals for pain monitoring
The method obtains at least two physiological signals, expressly including ECG plus at least one signal selected from a group comprising PPG, GSR, blood pressure, respiration, internal body temperature, skin temperature, EOG, pupil diameter, EEG, FEMG, EMG, EGG, partial pressure of carbon dioxide, and accelerometer readings.
Processing physiological signals for improved quality
The physiological signals are processed to improve signal quality, thereby producing processed signals that are suitable for feature extraction.
Generating a feature vector comprising specific extracted features
A first vector is generated comprising at least three features extracted from the physiological signals, where the features include at least one selected from RR/PQ/PR/QT/RS/ST intervals, their variability, Q/R/S/T/P amplitudes, or combinations thereof.
Transforming the feature vector with normalization
The first feature vector is transformed into a second vector, with the transformation including normalization steps to prepare data for classification.
Classifying pain status using an ensemble of classification and regression trees
The pain status of the patient is monitored by applying a classification algorithm adapted to classify the normalized second vector into a graduated scale representing pain level, wherein the classification algorithm comprises an ensemble of classification and regression trees.
The independent claim defines a comprehensive method combining acquisition, signal processing, feature extraction and transformation, and sophisticated classification, specifically using an ensemble of classification and regression trees, to monitor the pain status based on multiple physiological signals.
Stated Advantages
Improved pain and no-pain classification performance by combining multiple physiological signals compared to using single signals alone.
Capability to monitor pain in subjects across various states of consciousness including awake, semi-awake, and sedated.
Use of a large plurality of physiological features enhances the accuracy and robustness of pain detection and classification.
Adaptability to classify pain into multiple classes and scalable scoring correlating to subjective pain scales.
Capability to incorporate a priori data to improve classification efficiency.
Provision for remote, semi-remote, or local monitoring configurations facilitating flexibility in clinical environments.
Documented Applications
Monitoring and classifying pain in subjects undergoing various painful stimuli such as cold pressor test and heat pain test.
Pain detection and classification during different states of consciousness including awake, semi-awake, and sedated subjects.
Use in clinical pain management to assist caregivers in assessing pain levels non-invasively.
Integration with drug administration systems for automatic, semi-automatic, or manual adjustment of pain medication delivery.
Remote or telemedicine applications for pain monitoring via communication modules.
Classification of different pain types and intensities, such as severe pain versus mild pain, no pain, or mental stress.
Interested in licensing this patent?