DIGITAL IMAGE PROCESSING AND METABOLIC PARAMETER LINEARITY TO NON-INVASIVELY DETECT ANALYTE CONCENTRATION
- Paper number
IAC-19,A1,IP,12,x52280
- Author
Mr. Joseph Allen Jr., United States, University of North Dakota
- Year
2019
- Abstract
Spectroscopy is the scientific technique of quantifying and measuring electromagnetic, or light, reflectance or absorption. Atoms emit and/or absorb light when light passes through. The excitations provide specific energy signatures that relate to the element that is emitting or absorbing the light. Non-invasive biosensors monitor physical health properties such as heart rate, oxygen saturation, and tissue blood flow as a result of spectroscopy. Several attempts have been made to non-invasively detect metabolic chemical, or analyte, concentration (e.g. glucose) with various spectroscopic techniques. The primary limitation is due to signal-to-noise ratio. This research focuses on a unique method that combines emission spectroscopy and machine learning to non-invasively detect glucose and other metabolic analyte concentrations. Artificial neural network is applied to train a predictive model that enables the remote sensing capability. The data acquisition requires capturing digital images of the spectral reflectance. Image processing and segmentation determines discrete variables that correlate with the metabolic analytes. The clinical trial protocol includes n=90 subjects, and a venipuncture comprehensive metabolic panel blood test within two minutes prior to a non-invasive spectral reading. Results indicate a strong correlation between the spectral system and the clinical gold standard, relative to metabolic analyte concentration.
- Abstract document
- Manuscript document
(absent)