Optimization of Dual-Channel Near-Infrared Non-Invasive Glucose Level Measurement Sensors Based On Monte-Carlo Simulations

Abstract

Non-invasive glucose monitoring sensors are promising techniques in diabetes management. In particular, optical-based nearinfrared glucose sensors are label-free, compact, user-friendly, and inexpensive. They require no daily calibration and can provide continuous glucose level monitoring. However, these sensors are still in the development stage since their accuracy is still not widely accepted by medical professionals. Here, we introduce an optimized dual-channel approach for this kind of sensor where four optodes for short and long channels are employed. The long channel signal that contains useful information about the glucose level can be utilized to calculate the glucose contents in the dermis layer of the skin while the short channel is utilized to measure the interference "noise" signal originating from the epidermis layer. So, the latter can then be removed from the long channel signal. This work investigates the optimal source-detector separation (SDS) of both short and long channels for a wide range of wavelengths from 1200 nm to 1900 nm. Based on the sensitivity distributions of both dermis and epidermis layers and the signal-to-noise ratio, it is suggested that a 2 mm SDS distance with a wavelength of 1450 nm is a potential choice for the short channel. Moreover, 6 mm SDS at the wavelength of 1750 nm is the optimal choice for the long channel. This work may pave the way for further optimization of non-invasive near-infrared glucose sensors that will be widely adopted in the healthcare sector.

Is Raman the best strategy towards the development of non-invasive continuous glucose monitoring devices for diabetes management?

“…Accordingly, complex machine learning model and multivariate calibration models, such as partial least squares (PLS) regression, support vector regression or Monte-Carlo simulation, are required for extracting a quantification of glucose in the presence of other physiological substances and tissue components (water, haemoglobin, proteins, fat, etc.) ( Kino et al, 2016 ; Althobaiti and Al-Naib, 2021 ). For better understanding the state of the art of the non-invasively measured NIR signals from tissue, see studies from the Heise group, who reviewed the progress in emerging glucose monitoring techniques exploiting photoplethysmography within the visible and near-infrared range ( Delbeck et al, 2019 ; Heise et al, 2021 ).…”

In Silico Investigation of SNR and Dermis Sensitivity for Optimum Dual-Channel Near-Infrared Glucose Sensor Designs for Different Skin Colors

“…To systematically assess the performance when changing the wavelength and the SDS in order to choose the optimal NIR channel for measuring glucose content, we previously introduced [ 26 ] three metrics. Briefly, the first metric is the epidermis sensitivity, which is the summation of the photon density function (PMDF) for all voxels in the epidermis layer over the summation of all photon density functions (PMDF) in the model: …”

“…By running the MCS for multiple independent seeded simulations, one can calculate the mean (μ) and standard deviation ( σ ) at each voxel in the model. Thus, one can calculate the SNR (in decibels) as follows [ 26 ]: …”

“…In a previous study [ 26 ], we introduced an optimized NIR sensor with two channels for blood glucose measurements. The long channel is utilized to measure the glucose level in the inner skin (dermis) layer.…”

A Review of Non-Invasive Optical Systems for Continuous Blood Glucose Monitoring

“…The results were benchmarked with invasive devices for each volunteer and there was a good agreement between both methods. More recently, a dual-channel approach for NIR sensors is introduced where four optodes for short and long channels are utilized [25]. Figure 5 shows the proposed dualchannel near-infrared sensor.…”