Location
https://www.kennesaw.edu/ccse/events/computing-showcase/sp25-cday-program.php
Event Website
http://ebelfarsi.com/in_vitro
Document Type
Event
Start Date
15-4-2025 4:00 PM
Description
This study explores non-invasive glucose sensing using infrared (IR) imaging and electrical measurements in an in-vitro setup. Glucose samples (70–200 mg/dL) were prepared by diluting concentrated solutions (700–2000 mg/dL) 1:10 in synthetic blood concentrate, with 2 mg/dL increments. A custom 3D-printed black cuvette holder ensured consistent alignment of components, including either an IR camera or a 1550 nm photodiode, light sources (850 nm LED/laser, 808 nm, 650 nm, or 1600 nm), and a 3 mm skin-mimicking silicone layer. A Region Based Convolutional Neural Network (RCNN) trained on IR images achieved the lowest RMSE of 10.98 mg/dL at 850 nm LED. A Random Forest model using the recorded-to-baseline voltage ratio yielded an R² of 0.786, RMSE of 17.62 mg/dL, and MAE of 14.05 mg/dL. Clarke Error Grid analysis confirmed clinical relevance.
Included in
GRM-093 Advances in Non-Invasive Glucose Sensing: A Comprehensive In Vitro Analysis
https://www.kennesaw.edu/ccse/events/computing-showcase/sp25-cday-program.php
This study explores non-invasive glucose sensing using infrared (IR) imaging and electrical measurements in an in-vitro setup. Glucose samples (70–200 mg/dL) were prepared by diluting concentrated solutions (700–2000 mg/dL) 1:10 in synthetic blood concentrate, with 2 mg/dL increments. A custom 3D-printed black cuvette holder ensured consistent alignment of components, including either an IR camera or a 1550 nm photodiode, light sources (850 nm LED/laser, 808 nm, 650 nm, or 1600 nm), and a 3 mm skin-mimicking silicone layer. A Region Based Convolutional Neural Network (RCNN) trained on IR images achieved the lowest RMSE of 10.98 mg/dL at 850 nm LED. A Random Forest model using the recorded-to-baseline voltage ratio yielded an R² of 0.786, RMSE of 17.62 mg/dL, and MAE of 14.05 mg/dL. Clarke Error Grid analysis confirmed clinical relevance.
https://digitalcommons.kennesaw.edu/cday/Spring_2025/Masters_Research/15
