Machine learning-based identification of key electromyography and kinematic features for chronic neck pain classification

Jiménez-Grande, David ORCID: 0000-0001-5454-7667 (2024). Machine learning-based identification of key electromyography and kinematic features for chronic neck pain classification. University of Birmingham. Ph.D.

Preview

JimenezGrande2024PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (7MB) | Preview

Abstract

In the realm of neck pain management, researchers and medical professionals are constantly exploring new methods. This includes better assessments of pain and related mobility issues, as well as enhancing treatment to improve the recovery of individuals with neck pain. This thesis centres on the utilisation of machine learning (ML) techniques to investigate how electromyography (EMG) and kinematic features can be used to classify people with or without chronic neck pain as this may ultimately guide improved assessment and management of patients with neck pain disorders and associated movement impairments.

Specifically, the objective of this research is to develop a deeper understanding of how EMG signals and kinematic measurements differ between people with and without chronic neck pain and how such data can be used for classification purposes. By analysing and identifying patterns within these data streams, valuable insights can potentially be gained to aid in the assessment, and treatment of neck pain conditions.
To achieve this goal, a comprehensive literature review of relevant studies on neck pain assessment and ML applications was conducted (Chapter 1). Various EMG and kinematic datasets were then collected from participants with and without chronic neck pain, and appropriate feature extraction techniques were applied. ML algorithms were then employed to classify groups with and without neck pain and then, identify key EMG and kinematic features (Chapter 2). These methodologies were examined across diverse tasks, including dynamic contractions of the neck (Chapter 3), static posture (Chapter 4), and gait (Chapter 5).

The results of this thesis showed, firstly, the ability of different ML algorithms to accurately classify people with chronic neck pain compared to healthy individuals across a wide variety of tasks. Secondly, the results identified and highlighted key EMG and kinematic characteristics that improved the performance of all algorithms. These characteristics provide insights into potential muscle activity as well as movement anomalies that are present in people with chronic neck pain. These findings offer a significant step forward in the understanding of the biomechanical and neuromuscular differences in individuals with neck pain compared to pain-free individuals. Additionally, the identified EMG and kinematic characteristics can potentially serve as a foundation for the development of targeted rehabilitation protocols, aiming to address the specific neuromuscular and movement abnormalities found in patients with chronic neck pain. Future research can delve deeper into the causal relationships between these identified characteristics and the presence of neck pain, potentially leading to preventive strategies and interventions.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):