CAL neural network for precise severe obstructive sleep apnea detection

Abstract

Aim and Background: Obstructive sleep apnea (OSA) presents significant health risks and requires accurate detection for effective intervention. This study introduces the convolutional neural network with artificial neural network with long-term short memory (CAL) neural network, a novel approach that combines convolutional neural network (CNN), artificial neural network (ANN), and long-term short memory (LSTM) techniques to enhance OSA detection. The network leverages spatial feature extraction, pattern recognition, and temporal dependency modeling to improve detection accuracy. Methods: The proposed CAL neural network undergoes preprocessing, including median filtering for noise removal, before analyzing time-series representations of physiologica signals such as electroencephalogram, electrocardiogram, and respiratory signals. The CNN component extracts spatial features from raw signals, capturing relevant patterns and relationships among different channels. The output of the CNN is then integrated into an ANN layer to refine learned features and identify complex patterns indicative of severe OSA. LSTM layers are subsequently introduced to capture long-term dependencies in signal dynamics, crucial for detecting subtle variations associated with severe OSA events. Results: Preliminary results demonstrate the effectiveness of the CAL neural network in detecting severe OSA. By combining CNN, ANN, and LSTM techniques, the network achieves improved accuracy in identifying OSA-related patterns and variations in physiological signals. Conclusion: The CAL neural network presents a promising approach for OSA detection, offering enhanced accuracy through the synergistic integration of CNN, ANN, and LSTM techniques.