This paper presents an improvement of classification performance for electroencephalography (EEG)-based driver fatigue classification between fatigue and alert states with the data collected from 43 participants. The system employs autoregressive (AR) modeling as the features extraction algorithm, and sparse-deep belief networks (sparse-DBN) as the classification algorithm. Compared to other classifiers, sparse-DBN is a semi supervised learning method which combines unsupervised learning for modeling features in the pre-training layer and supervised learning for classification in the following layer. The sparsity in sparse-DBN is achieved with a regularization term that penalizes a deviation of the expected activation of hidden units from a fixed low-level prevents the network from overfitting and is able to learn low-level structures as well as high-level structures. For comparison, the artificial neural networks (ANN), Bayesian neural networks (BNN), and original deep belief networks (DBN) classifiers are used. The classification results show that using AR feature extractor and DBN classifiers, the classification performance achieves an improved classification performance with a of sensitivity of 90.8%, a specificity of 90.4%, an accuracy of 90.6%, and an area under the receiver operating curve (AUROC) of 0.94 compared to ANN (sensitivity at 80.8%, specificity at 77.8%, accuracy at 79.3% with AUC-ROC of 0.83) and BNN classifiers (sensitivity at 84.3%, specificity at 83%, accuracy at 83.6% with AUROC of 0.87). Using the sparse-DBN classifier, the classification performance improved further with sensitivity of 93.9%, a specificity of 92.3%, and an accuracy of 93.1% with AUROC of 0.96. Overall, the sparse-DBN classifier improved accuracy by 13.8, 9.5, and 2.5% over ANN, BNN, and DBN classifiers, respectively.

No abstract

Hypoglycemia (low blood glucose level) is a medical emergency and is a very common in type 1 diabetic persons and can occur at any age. It is the major limiting factor to maintain tight glycemic control. The deficiency in glucose counter-regulation may even lead to severe hypoglycaemia. It is always threatening to the well-being of patients with Type 1 diabetes mellitus (T1DM) since more severe hypoglycemia leads to seizures or loss of consciousness and the possible development of permanent brain dysfunction under certain circumstances . Because of that, an accurate continuing hypoglycemia monitoring system is a very important medical device for diabetic patients. Traditionally, different electrochemical glucose meters are adopted to measure the blood glucose. However, they are invasive, discontinuing and uncomfortable methods or painful for patients. In this paper, we proposed a noninvasive hypoglycemia monitoring system using the physiological parameters of electrocardiography (ECG) signal such as heart rate and corrected QT intervals. To enhance the detection accuracy of the hypoglycemia, a new emerging technology called extreme learning machine (ELM) is developed to recognize the presence of hypoglycemic episodes. ELM have both universal approximation and classification capabilities and provides efficient unified solutions to generalize feed-forward neural networks. A real clinical study of sixteen children with T1DM are given in this paper to illustrate the performance of ELM and successfully applied to a non-invasive hypoglycemia monitoring system. The natural occurrence of nocturnal hypoglycemic episodes are associated with increased heart rates (p < 0.06) and increased corrected QT intervals (p < 0.001). For comparison purpose, different computational intelligence technologies such as swarm based neural network, multiple regression based-fuzzy inference system, and fuzzy system were applied to this hypoglycemia monitoring system. By using the proposed ELM-based neural network (ELM-NN), the testing sensitivity (true positive) and specificity (true negative) for detection of hypoglycemia is better than the other intelligence methods with faster training time.