Princeton University

School of Engineering & Applied Science

Development of a Noninvasive IN VIVO Glucose Sensor Based on Mid-Infrared Quantum Cascade Laser Spectroscopy

Sabbir Liakat
Engineering Quadrangle B327
Tuesday, April 21, 2015 - 3:00pm to 4:30pm

Diabetes mellitus impacts over 340 million people across the world. Diabetics must monitor their glucose levels multiple times a day and currently, the most accurate method of doing so involves pricking one’s finger, an often painful procedure.  This dissertation documents research aimed at developing a noninvasive in vivo glucose sensor, with the motivation of implementing a pain-free means of measuring glucose for diabetics.  
We utilize a novel approach based on mid-infrared (mid‐IR) Quantum Cascade (QC) laser spectroscopy. While previous optical solutions using near‐IR light between 1300-1900 nm have failed due to limits on glucose absorption feature specificity, the mid-IR region between 8 – 10 μm contains strong absorption features with cross-sections up to four orders of magnitude greater than their near-IR counterparts.  Although mid-IR spectroscopy has traditionally been neglected for in vivo applications due to the lack of light sources capable of sufficient skin penetration, the advent of powerful QC lasers allows us to overcome this limitation. Here, theoretical background of light-matter interaction is presented, followed by discussion of experimental research progressing towards noninvasive glucose sensing.
First, the feasibility of mid-IR noninvasive glucose sensing is shown through measurements of angular scattering patterns in skin, which show that QC laser light can penetrate deep enough into skin.
Next, we show clinically accurate sensing of physiological glucose concentrations in vitro using partial least squares regression (PLSR) analysis on mid-IR transmission spectra of proxy solutions for dermal interstitial fluid.
Finally, we record breakthrough results showing clinically accurate glucose prediction capability in vivo with human subjects using a setup featuring hollow-core fiber optics, liquid nitrogen cooled detection, and PLSR analysis. This sensor is then made more compact by replacing the fibers with an integrating sphere that significantly increases the collection efficiency of scattered light from skin. This allows the sensor to be housed on a mobile cart, and it removes the system’s nitrogen dependency by allowing the use of a thermoelectrically cooled detector, all while maintaining glucose sensing accuracy.  We conclude with an outlook for the improved portable sensor to move on from laboratory trials with volunteers to large-­‐scale trials in diabetes clinics.