Deep Artificial Denoising Auto-Encoder-Decoder Pain Recognition System

The opportunity to develop low-cost, accurate and automatic pain recognition systems that utilizes low resolution ECG data to improve pain recognition is vital for healthcare and medical research. Current works on pain analysis use high resolution equipments to collect data and rely on patients'...

Full description

Saved in:
Bibliographic Details
Published in:Proceedings (International Conference on Machine Learning and Cybernetics.) pp. 75 - 81
Main Authors: Fordson, H. Perry, Anderson, Adam, Derosa, Eve
Format: Conference Proceeding
Language:English
Published: IEEE 09.07.2023
Subjects:
DADAED
ECG
Electrocardiography
Emotion
Industries
Medical services
Mental health
Noise reduction
Pain
Pain recognition
Pain threshold
Tolerance
ISSN:2160-1348
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The opportunity to develop low-cost, accurate and automatic pain recognition systems that utilizes low resolution ECG data to improve pain recognition is vital for healthcare and medical research. Current works on pain analysis use high resolution equipments to collect data and rely on patients' subjective pain sensation. This mostly ignores individuals interoceptive accuracy which can be unreliable and impractical especially when people cannot verbally report on their pain. In this paper, we propose a deep neural network that incorporates an artificial denoising auto-encoder-decoder system to automatically generate and select applicable signifiers directly from ECG data. We also extract state of the art statistical, inter-beat interval, heart rate variability, and heart rate features from the data. We then concurrently optimize both manually extracted features and the Deep Artificial Denoising Auto-Encoder-Decoder (DADAED) features and perform a binary classification of pain threshold and pain tolerance. Our work is finally assessed on a BioVid Heat Pain Database (Part B). The simultaneous optimization approach of the deep neural techniques introduces trainable weighted parameters and generates latent representation. This outperforms previously proposed methods on the same database and other related works on pain recognition. Our pain recognition system has the potential to revolutionize the pain management industry, emotion regulation industry, and mental health centers.
ISSN:2160-1348
DOI:10.1109/ICMLC58545.2023.10327969