Dealing with Imbalanced Sleep Apnea Data Using DCGAN

Data in the health sector are often lacking and unbalanced. It is because collecting data takes time many resources. One example sleep apnea which about 8–10 hours to get uses specialized hardware like polysomnography (PSG). This study proposes a augmentation technique handle unbalanced using DCGAN several deep learning models such as 1D-CNN, ANN, LSTM, 1D-CNN+LSTM classifier for detection. The architecture used CNN on generator discriminator. will create new synthetic by mimicking original dataset. experiment dataset from PhysioNet, Apnea-ECG, MIT-BIH PSG Database. Furthermore, preprocessed remove noise, features extracted manually. test scenario 10% 50% be added Then compare performance of multiple before after adding data. results indicate that with can improve almost all models, highest increase 1.78% model 4.80% LSTM Apnea-ECG datasets, respectively.