Thompson, S (2025) Detection of Obstructive Sleep Apnoea using Features Extracted from Segmented Time-Series ECG Signals with a One Dimensional Convolutional Neural Network. Doctoral thesis, Liverpool John Moores University.

Preview

Text 2025ThompsonPhD.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (2MB) | Preview

Abstract

The study in this paper presents a one-dimensional convolutional neural network (1DCNN) model, designed for the automated detection of obstructive Sleep Apnoea (OSA) captured from single-channel electrocardiogram (ECG) signals. This system presents a novel idea that could provide support mechanisms in clinical practice to promptly diagnose patients suffering from OSA. Using the state-of-the-art in 1DCNNs, a model is constructed using 4 major parts, a convolutional feature layer, max pooling layer, activation layer and a fully connected Multilayer Perceptron (MLP), comprising of a hidden layers and a SoftMax output for classification. The model repeatedly learns how to extract prominent features from the one-dimensional data before mapping it to the MLP to better learn complex feature relationships and increase classification. To improve training the novel idea to produce specific window sizes of time-series ECG data was introduced. For training and validation of the model, 35 ECG signal recordings were selected from an annotated database containing 70 night-time ECG recordings. (Group A = a01 to a20 (Apnoea breathing), Group B = b01 to b05 (moderate), and Group C = c01 to c10 (normal). Performance of the model was evaluated by using various metrics. For the testing stage 35 recording are withheld. Test results showed the 1DCNN-500 model produced excellent classification results with averaged scores of sensitivity 0.9742, specificity 0.9715, accuracy 0.9728. To further evaluate the ability of the IDCNN, the same datasets and training were applied to two alternative Machine Learning (ML) classification algorithms, namely the Random Forest Classifier (RFC) and the Support Vector Machine (SVM). The RFC produced scores of (Sensitivity/Recall (0) 0.90 / (1) 0.94, Precision (0) 0.94 / (1) 0.90, Accuracy 0.91) and the SVM produced scores of (Sensitivity (0) 0.94 / (1) 0.50, Precision (0) 0.65 / (1) 0.90, Accuracy 0.72). Analysing the results from all the experiments confirmed the IDCNN model can identify the presence of Apnoea with a higher degree of accuracy and rapidity than other traditional ML models.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Data Science; 1DCNN; one-dimensional convolutional neural network; Sleep Apnoea; Machine Learning; ML
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Computer Science and Mathematics
SWORD Depositor:	A Symplectic
Date Deposited:	04 Feb 2025 15:00
Last Modified:	04 Feb 2025 15:00
DOI or ID number:	10.24377/LJMU.t.00025248
Supervisors:	Reilly, D, Fergus, P and Shaw, A
URI:	https://researchonline.ljmu.ac.uk/id/eprint/25248

View Item