Pain-Affect Dataset Merging for Neural Network Pain Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated pain assessment methods are limited in real-world settings as they only compare pain to a neutral state, failing to account for multiple affect states experienced by patients, such as anxiety, depression, and comfort, which are common in clinical settings.
Innovation Solution
A system and method that incorporates affect states into pain assessment using a new pain-affect dataset, merging pain and affect datasets to train a neural network model for identifying pain levels, which includes image data and biopotential signals from patients, allowing for robust pain assessment in real-world settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing pain assessment models trained only on pain and neutral states are used, then standardized assessment and continuous monitoring are provided, but the models fail to accurately assess pain in real-world settings where patients experience multiple affect states
Solution Approach 1:
The patent merges the pain dataset and affect dataset into a unified pain-affect dataset, combining multiple modalities (image data, EDA, ECG, EMG) to train a single neural network model that can simultaneously assess both pain levels and affect states, thereby improving reliability across diverse clinical scenarios
Solution Approach 2:
The model transitions from a static classification approach (pain vs. neutral) to a dynamic multi-state classification system that can adapt to and distinguish between multiple affect states (anxiety, depression, comfort, etc.) in real-time, enhancing both accuracy and adaptability
2Ease of manufacture
If pain assessment models are trained only on pain and neutral states, then the assessment is simplified and standardized, but it cannot account for affect states such as anxiety, depression, and comfort that patients experience in clinical settings
Solution Approach 1:
The patent segments the training process into distinct phases: first training on the pain dataset, then training on the affect dataset, and finally merging them into a unified model. This segmented approach maintains training organization and simplicity while ensuring comprehensive coverage of both pain and affect states
Solution Approach 2:
The patent creates a composite dataset structure that integrates multiple data types (image, EDA, ECG, EMG) from both pain and affect studies, forming a rich multi-modal pain-affect dataset that preserves diverse information while maintaining a unified training framework
Data Source
AI summary
A system and method for pain level recognition using an automated approach which incorporates a pain-affect dataset comprising bioVid pain and bioVid emotion datasets for the assessment of patient pain in clinical settings where patients often experience other affect states, such as anger and anxiety, in addition to pain.