Empirical wavelet decomposition of photoplethysmographic signal for hypertension risk stratification and detection of diabetes mellitus using machine learning techniques

Hypertension (HT) is a leading risk factor for cardiovascular disease (CVD), and the overlap of diabetes mellitus (DM) with hypertension can lead to severe complications.Presently, the diagnostic method for detecting hypertension and DM is unsuitable for large-scale screening.The proposed model uses a statistical feature extracted by decomposing the PPG signal into a sub-band signal using empirical wavelet transform (EWT), a comparative study conducted between various soft and hard computing classification models.The highest accuracy achieved by sequential neural network for the three categories, namely normal (NT) vs. prehypertension (PHT), NT vs. hypertension type 1 (HT-I), NT vs. hypertension type 2 (HT-II) in terms of F1 scores is 78.9%, 91.2% and 94%, respectively and F1 score of 97.9% for detection of DM-II patients.We conclude that soft computing techniques such as deep learning neural networks have shown superior performance compared to hard computing techniques.Furthermore, features selected using a hybrid feature selection technique were found to improve the classifier's performance.The main advantage of the proposed model that uses a decomposition technique is found to be more immune to noisy PPG signals, overcoming the limitation of the morphological-based model.