Microcontroller Implementation of Support Vector Machine for Detecting Blood Glucose Levels Using Breath Volatile Organic Compounds.

This paper presents an embedded system-based solution for sensor arrays to estimate blood glucose levels from volatile organic compounds (VOCs) in a patient's breath. Support vector machine (SVM) was trained on a general-purpose computer using an existing SVM library. A training model, optimized to achieve the most accurate results, was implemented in a microcontroller with an ATMega microprocessor. Training and testing was conducted using artificial breath that mimics known VOC footprints of high and low blood glucose levels. The embedded solution was able to correctly categorize the corresponding glucose levels of the artificial breath samples with 97.1% accuracy. The presented results make a significant contribution toward the development of a portable device for detecting blood glucose levels from a patient's breath.

Analyzing breath samples of hypoglycemic events in type 1 diabetes patients: towards developing an alternative to diabetes alert dogs.

Diabetes is a disease that involves dysregulation of metabolic processes. Patients with type 1 diabetes (T1D) require insulin injections and measured food intake to maintain clinical stability, manually tracking their results by measuring blood glucose levels. Low blood glucose levels, hypoglycemia, can be extremely dangerous and can result in seizures, coma, or even death. Canines trained as diabetes alert dogs (DADs) have demonstrated the ability to detect hypoglycemia from breath, which led us to hypothesize that hypoglycemia, a metabolic dysregulation leading to low blood glucose levels, could be identified through analyzing volatile organic compounds (VOCs) contained within breath. We hoped to replicate the canines' detection ability and success by analytically using gas chromatography/mass spectrometry of VOCs in 128 breath samples collected from 52 youths with T1D at two different diabetes camps. We used different tests for significance including Ranksum, Student's T-test, and difference between means, and found a subset of 56 traces of potential metabolites. Principle component and linear discriminant analysis (LDA) confirmed a hypoglycemic signature likely resides within this group. Supervised machine learning combined with LDA narrowed the list of likely components to seven. The technique of leave one out cross validation demonstrated the model thus developed has a sensitivity of 91% (95% confidence interval (CI) [57.1, 94.7]) and a specificity of 84% (95% CI [73.0, 92.7]) at identifying hypoglycemia. Confidence intervals were obtained by bootstrapping. These results demonstrate that it is possible to differentiate breath samples obtained during hypoglycemic events from all other breath samples by analytical means and could lead to developing a simple analytical monitoring device as an alternative to using DADs.

Cross-Selectivity Enhancement of Poly(vinylidene fluoride-hexafluoropropylene)-Based Sensor Arrays for Detecting Acetone and Ethanol.

Two methods for cross-selectivity enhancement of porous poly(vinylidene fluoride-hexafluoropropylene) (PVDF-HFP)/carbon black (CB) composite-based resistive sensors are provided. The sensors are tested with acetone and ethanol in the presence of humid air. Cross-selectivity is enhanced using two different methods to modify the basic response of the PVDF-HFP/CB sensing platform. In method I, the adsorption properties of PVDF-HFP/CB are altered by adding a polyethylene oxide (PEO) layer or by treating with infrared (IR). In method II, the effects of the interaction of acetone and ethanol are enhanced by adding diethylene carbonate (DEC) or PEO dispersed in DEC (PEO/DEC) to the film. The results suggest the approaches used in method I alter the composite ability to adsorb acetone and ethanol, while in method II, they alter the transduction characteristics of the composite. Using these approaches, sensor relative response to acetone was increased by 89% compared with the PVDF-HFP/CB untreated film, whereas sensor relative response to ethanol could be decreased by 57% or increased by 197%. Not only do these results demonstrate facile methods for increasing sensitivity of PVDF-HFP/CB film, used in parallel they demonstrate a roadmap for enhancing system cross-selectivity that can be applied to separate units on an array. Fabrication methods, experimental procedures and results are presented and discussed.