Continuous glucose monitoring devices allow diabetes condition management. However, when limited data is available, one option is to increase their size by generating synthetic samples. From a homemade wearable prototype was created a real dataset with 18 instances and 53 attributes that capture characteristics of capillary and venous blood glucose, oxygen concentration, pulse rate, skin temperature, and 24 modules and 24 phases related to bio-impedance. The objective of this article is to generate synthetic datasets, and also it investigates the ideal features subset and optimal model for non-invasive diabetes prediction. Gaussian-Copulas (GC), conditional generative adversarial networks (CG), variational autoencoders, and Copula-GAN techniques' were used to generate five synthetic datasets. Experiments show that GC1 and GC2 datasets follow min/max boundaries and are not copies of the original data. Multilayer perceptron regressor outperformed (train and test) with 2.17, 2.51 in MAE; 9.29, 13.59 in MSE; 3.05, 3.69 in RMSE, and 0.95, 0.92 in R2 in GC1, and 2.64, 3.02 in MAE; 11.43, 15.11 in MSE; 3.38, 3.89 in RMSE, and 0.94, 0.92 in R2 in GC2 with eight features. Future work is necessary to explore autoencoder and generative architectures, datasets with diverse characteristics, and the effect of the number of features.

Chronic non-communicable diseases (NCDs) are major public health problems and a significant financial burden on public health systems. By 2030, mortality due to NCDs, such as cardiovascular diseases, cancer, respiratory diseases, and diabetes, is predicted to increase in Brazil. A peculiar aspect of NCDs involves their long-term and integrated care management. This paper proposes a chronic disease management platform based on artificial intelligence to deliver digital health services everywhere. The proposed platform is anchored and built with healthcare 4.0 technologies, such as wearable devices, the internet of medical things, and artificial intelligence cloud-based solutions that allow the deployment of a smart healthcare system. In addition, the paper presents the feasibility of the platform in a diabetes prediction study case. For the study case, an initial dataset was established with bio-impedance, oxygen concentration, pulse rate, skin impedance, and skin temperature attributes. A baseline was implemented with ten regression models to assess the prediction performance of the mean squared error, root mean squared error, and r-squared score to compare predictive findings with capillary blood glucose measurements. Results evidence that the decision tree regressor and three ensemble methods (bagging decision tree regressor, random forest regressor, and AdaBoost regressor) yielded improvement over the other models. Moreover, a comparison among those models revealed that the decision tree regressor outperforms them and presents promissory outcomes.

Pathogens and adulterants in human feeding consumables can be readily identified according to their electrical properties. Electrical bioimpedance analysis (BIA) has been widely used for body contents characterization, such as blood, urine, lactate and sweat. If the blood glucose concentration alters the electrical properties of the blood medium, then the impedance spectrum obtained by BIA can be used to measure glycemia. In some applications, artificial neural networks allows the correlation of these parameters (impedance and glucose concentration) by means of symbolic and statistical rules. According to our literature review, there is any physical model that allows the interpretation of the relationship between blood’s electrical properties, obtained by BIA, and the concentration of glucose in the blood plasma. This article proposes a simplified physical model for blood electrical conductivity as a function of glucose concentration, based on Bruggeman’s effective medium theory. The equations of this model were obtained considering an insulating phase distribution diffused in a conductive matrix, in which red blood cells are represented by macroscopic insulating nuclei and glucose molecules by microscopic insulating particles. The impedance spectrum for different glucose concentrations (4.0 to 6.8 mmol/L) in a blood sample, published by Kamat Bagul (2014), were compared with the proposed model. The results showed a significant correlation with the experimental data, showing a maximum error of 5.2%. The proposed model might be useful in the design of noninvasive blood glucose monitoring systems.

Parasitic capacitance represents the main error source in measurement systems based on electrical impedance spectroscopy. The capacitive nature of electrodes’ impedance in tetrapolar configuration can give origin to phase errors when electrodes are coupled to parasitic capacitances. Nevertheless, reactive charges in tissue excitation systems are susceptible to instability. Based on such a scenario, mitigating capacitive effects associated with the electrode is a requirement in order to reduce errors in the measurement system. A literature review about the main compensation techniques for parasitic capacitance was carried out. The selected studies were categorized into three groups: (i) compensation in electronic instrumentation; (ii) compensation in measurement processing, and (iii) compensation by negative impedance converters. The three analyzed methods emerged as effective against fixed capacitance. No method seemed capable of mitigating the effects of electrodes’ capacitance, that changes in the frequency spectrum. The analysis has revealed the need for a method to compensate varying capacitances, since electrodes’ impedance is unknown.

Current sources play an essential role in tissue excitation used in bioelectrical impedance spectroscopy. Most investigations use Howland current sources that, despite their practicality and simplified implementation, have operating frequency limitations and dependence on the load impedance due to theirs narrow output impedance, specially at higher frequencies. The objective of this work is to propose model for a robust current-controlled sinusoidal oscillator. The oscillator is based on fully analog electronics, which enables controlling oscillation phase and amplitude by using a voltage reference. The mathematical model is based on Pyragas control application to the classical harmonic oscillator. From the modelling process, it was build an oscillator topology based on second-generation current carriers and on transconductance amplifiers. The reference signal (Fsync) was a sinusoidal voltage source having a frequency of 1MHz and an amplitude of 1Vpp. The oscillator output current synchronized the oscillations’ phase and amplitude with Fsync, regardless of their magnitude before the control signal acted in the circuit at t≈13.5μs. SPICE simulations using ideal components have confirmed the successful operation of the proposed oscillator. This type of oscillator can be implemented in SOIC, then allowing oscillation control interface with logic circuits.