Obstructive Sleep Apnea (OSA) poses significant health risks and necessitates accurate detection for effective intervention. In this study, we introduce a novel approach, the CAL Neural Network, which synergistically combines the strengths of Convolutional Neural Network (CNN), Artificial Neural Network (ANN), and Long-Term Short Memory (LSTM) techniques to enhance the accuracy of severe OSA detection. The dataset has collected from Kaggle repository, undergoes preprocessing involving median filtering for noise removal. Our proposed CAL Neural Network exploits the spatial feature extraction prowess of CNNs, the pattern recognition capabilities of ANNs, and the temporal dependencies modeling of LSTMs. Time-series representations of physiological signals, including electroencephalogram (EEG), electrocardiogram (ECG), and respiratory signals, constitute the input data. The CNN component automatically extracts spatial features from raw signals, capturing pertinent patterns and relationships among different channels. The output of the CNN is then seamlessly integrated into an ANN layer, refining learned features and discerning complex patterns indicative of severe OSA. Subsequently, LSTM layers are introduced to capture long-term dependencies in the temporal dynamics of input signals, essential for detecting subtle variations associated with severe OSA events.
Key words: Artificial Neural Network, Convolutional Neural Network, CAL Neural Network,
Median Filtering, Obstructive Sleep Apnea
|
