I chose to focus on glucose monitoring for diabetics due to the increasing number of diabetes cases in Indonesia and the lack of relevant advanced technology, including glucometers, in local healthcare facilities. Common invasive glucometers available are expensive and uncomfortable due to the need to draw blood.
After looking at previous trials and studies, I selected three methods – thermal technology, reverse iontophoresis, and optical coherence tomography – to test out. Amongst these three, thermal technology and optical coherence tomography were the most accurate. Before creating a final design, I modified these two mechanisms, in the process turning the OCT method into an interferometric system.
I then hypothesised that if interferometry and thermal technology are utilised in the measurement of blood glucose concentration, with resistance of a light-dependent resistor and thermistor respectively used as indicators, accurate glucose concentration readings can be obtained in a non-invasive and continuous, yet affordable method.
The thermal technology uses the correlation of heat capacity of the epidermis to glucose concentration and measures the rate of change of heat through a thermistor. The interferometric system relies on the correlation between the ratio of the refractive indices of the interstitial fluid and scattering centres.
My hypothesis was supported, with the product achieving an R2 value of 0.843 through the usage of multiple sensors whilst costing only about $63. In the future, I would like to extend the range of substances in the blood this device can measure and market it to the public.
Primary Question: "Can we accurately and continuously track glucose levels in the body without having to withdraw blood?"
Secondary Question: "Can a noninvasive continuous glucometer with interferometry and thermal technology, which both utilise resistance as a marker for glucose concentration, be developed to give out accurate readings while still being affordable to the general Indonesian public?"
Currently, there are no noninvasive continuous glucose monitors available to the public. The next best options – continuous invasive monitors and intermittent noninvasive monitors – are expensive, retailing for at least $1,000. As a result, many people in Indonesia are not able to monitor glucose levels in an accurate and comfortable manner, if at all. This is proving itself a problem due to the increasing number of diabetics yearly.
In order to combat this issue, I have decided to develop a noninvasive continuous glucose monitor that is affordable and accurate. This is done through the usage of multiple indicators in one product simultaneously measuring glucose concentration, namingly systems utilising thermal technology and interferometry.
My hypothesis is as follows:
“if interferometry and thermal technology are utilised in the measurement of blood glucose concentration, with resistance of a light-dependent resistor and thermistor respectively used as indicators, accurate glucose concentration readings can be obtained in a non-invasive and continuous, yet affordable method.”
Diabetes in Indonesia: Statistics
There were over 10 million cases of diabetes in Indonesia in 2017 alone, 70% of which were undiagnosed and consequently left untreated. As the fourth leading cause of death, with a 6.7% prevalence in adults, diabetes has earned a name as a deadly killer in this archipelago. Despite so, available facilities are scarce and technologically behind, with even insulin being generally unavailable in healthcare centres.
According to the World Health Organization (2016), the only form of diagnostic measures commonly present in primary care facilities is a blood glucose measurement test. Home measurement kits, on the other hand, are found in drugstores across more developed cities, but often utilises outdated technology that limits accuracy.
Limitations in Glucose Monitoring
As of 2018, there are no noninvasive constant glucose trackers available to the general public. The only currently available noninvasive glucose tracker, GlucoTrack, utilises three different methods (thermal, electric, and ultrasonic). Although this method has been proven effective, with 96% of clinical trial results falling into the A and B zones of a Clarke error grid, it is unable to continuously measure glucose levels. As a result, users are not informed of potentially hazardous undetected highs and lows.
Formerly released commercial noninvasive devices, such as the GlucoWatch, have been reported to be inaccurate and create irritation and scarring on the user's skin due to prolonged exposure.
Previous Research in Noninvasive Glucose Monitoring
Amongst all noninvasive methods previously researched on and developed by scientists over the past 20 years, I have selected three different mechanisms to further inspect and directly evaluate. These mechanisms have been chosen due to their unique approach in monitoring glucose levels and their relative success in clinical trials:
This research has made me aware of the extent to which Indonesia is affected by diabetes mellitus. In addition to this, I have discovered the lack of technology across the country and the demand for constant monitoring supplies and consequently formulated the idea of creating an affordable noninvasive monitor for the public to access.
I have also examined and analysed multiple mechanisms used in less successful commercially developed trackers and selected three to directly develop and test, two of which will be incorporated in my final design.
Firstly, I tested out three techniques aforementioned in the research section (thermal technology, reverse iontophoresis, and optical coherence tomography) to assess the accuracy of each mechanism.
I took ten readings per mechanism and plotted these readings in graphs against blood glucose concentration given by an invasive glucose monitor. The R2 value received (through the least squares method) acted as the dependent variable.
The two most accurate mechanisms will be incorporated in my final product.
I measured the rate of temperature change and calculated the equivalent thermal signal:
Tear - temperature of ear;
Tamb - ambient temperature;
k - temperature correction factor;
tf - finishing time;
t0 - starting time.
This was done through using the apparatus below:
This experiment was done with reference to the GlucoTrack device and its testing.
Iontophoresis utilises a voltage gradient across the skin to extract molecules – in this case, glucose and lactate, which are later separated in a microcentrifuge. The glucose-to-lactate ratio is measured against the true blood glucose concentration.
The apparatus below is used to receive an OCT image, from which a slope dependent on the refractive indices acquired can be calculated:
This experiment was done with reference to this article.
Changes made to the two most accurate mechanisms (OCT and thermal technology) in the final product are elucidated on below:
These mechanisms are then connected to a microcontroller which triggers the closing of a switch that closes the circuit and allows a current to flow. This closing was programmed to occur in intervals of 5 minutes.
These new mechanisms are tested 12 times each and their corresponding signals plotted in a graph against true glucose concentration measured by an invasive glucometer. In thermal technology, resistance of the thermistors after a set period of time were measured, while resulting resistance in a light-dependent resistor (LDR) was measured in the OCT/interferometry system.
The final product's readings were acquired through this method:
Readings from my product were taken in 5-minute intervals, similarly to other continuous monitors, in an 8-hour period to produce 96 data points. These points are measured against readings given by an invasive glucose monitor and plotted in a graph.
To measure accuracy, the R2 value of the graph will be calculated. Furthermore, the resultant points will be plotted on a Clarke error grid and the percentage of points in medically acceptable zones evaluated.
The entirety of experiments throughout this project was conducted under room temperature. All graphs include at least 10 readings to enable precise statistical analysis. Additionally, the same glucometer was used throughout.
In order to ensure safety, all electrical and thermal components are monitored under adult supervision. Furthermore, glucose strips and needles used with the invasive glucometer are new and single-use.
Out of the three mechanisms I observed and tested, the two most accurate were thermal technology (R2 = 0.833) and optical coherence tomography (R2 = 0.778). The third mechanism, reverse iontophoresis, resulted in an R2 value of 0.601 which increased to 0.746 upon the removal of an outlier (circled). Consequently, these two methods will be incorporated in my final product and developed to produce more accurate results.
The graphs of respective signals against true blood glucose concentration. From top to bottom: thermal technology, reverse iontophoresis, optical coherence tomography.
The two most accurate mechanisms, thermal technology and OCT, were modified. These improved mechanisms were then tested out a further 12 times to evaluate whether these modified mechanisms were more accurate than the initial mechanisms. In addition to this, the logarithmic trend lines obtained will be used in the final product to obtain readings.
The thermal technology modified mechanism saw an increase in R2 value from 0.833 to 0.853. This indicates that the modified mechanism is more accurate and effective than the initial mechanism referred to in the beginning of the project.
The logarithmic trend line obtained can be expressed as a function T(x) = -3.59 ln(x) + 28.01.
The interferometry/OCT mechanism, on the other hand, showed a decrease in R2 value from 0.778 to 0.755. Although this suggests a decline in accuracy, I would like to still consider this modification a success due to the large amount of costs cut through the adjustments made. Although the R2 value does not meet the accuracy threshold of 0.8, it is likely that the values from the two mechanisms will 'balance' each other out when averaging, resulting in an accurate final product.
The logarithmic trend line obtained can be expressed as a function I(x) = -383.94 ln(x) + 2448.09.
In both mechanisms in the final product, resistance was measured and mapped into their corresponding functions (mentioned in method, section 3) to receive estimates for glucose concentration. The two values obtained are averaged out to give the final reading. 96 data points, with range 100, were obtained.
In a graph of measured glucose concentration (given by my product) against true glucose concentration (given by an invasive glucose monitor), the R2 value obtained was 0.843, suggesting a strong positive correlation. This confirms my hypothesis that a noninvasive method of continuous glucose monitoring can be achieved at a high degree of accuracy.
To further confirm the accuracy of the final product, I plotted these 96 points on a Clarke error grid (typically used when testing diabetes-related technology). A Clarke error grid is divided into five zones:
95 out of 96 points (~99%) were in the clinically accepted A + B zones and would result in positive medical decisions.
This study has confirmed that "an accurate and affordable device that can measure blood glucose concentration in a noninvasive and continuous manner can be created". The method demonstrated achieved a coefficient of determination of 0.843 and 99% of points plotted were in the clinically available A+B zones of a Clarke error grid, despite only costing about $63 per unit. The utilisation of two indicators in my product helped rule out anomalies and balance out extremities, hence improving accuracy. Hence, my hypotheses are accepted.
The final product solves the problems with glucose monitors posed (that many Indonesians are unable to efficiently measure blood glucose concentration due to three factors – pain as a result of invasive techniques, the intermittent nature of affordable monitors, and the high costs of painless/continuous monitors). The ability of this product to measure glucose continuously in a painless manner allows diabetics to monitor levels more closely, hence allowing them to make decisions that will help maintain their health.
I believe that this experiment has been highly successful. This is because I have developed an unprecedented model utilising two vastly different scientific approaches to attain highly accurate results. In addition to this, this experiment was closely and effectively controlled, with little room for error. Thus, the data points obtained are valid and have small error margins.
The limitations I encountered during this project are:
I will continue to pursue this project in the near future. I am planning to explore these potential aspects:
Hi, I’m Celestine! I’m a high school sophomore in the British School Jakarta, in Indonesia. I am also a sciences enthusiast, an avid reader, a bubble tea expert, and most of all, a dreamer.
What makes research so appealing to me is how one unsolved problem can be approached in a multitude of methods. As a seeker of perception and insight, I find experimenting inspiring as it gives me new visions of the world that cannot be found in textbooks or syllabi.
My influences include not only scientists and enterpreneurs, like Musk and Hawking, but suffragettes and advocates who strive for societal good. A scientist I particularly look up to is Christiaan Eijkman, a Nobel Prize recipient who has experimented on vitamins in my home country.
Winning the Google Science Fair – one of the most prestigious competitions in STEM for high-schoolers – would create new frontiers and pathways for my future in the sciences. It would enable access to opportunities, both for myself and my research, previously inconceivable. Living in a third world country, technology and research facilities are scarce, thus narrowing my options for development and scientific work.
In addition to this, I’d like to use this opportunity to convey a message to kids and teenagers across the world: that it is not too early to start working towards your hopes and ambitions. That if some girl can develop an award-winning device, you can make a change, too.
The entirety of my manufacturing process was performed in the technology workshop of the British School Jakarta:
Technology Workshop, British School Jakarta
Jalan Ciledug Raya, Bintaro Sektor 9, Jl. Jombang Raya,
Parigi, Pd. Aren, Kota Tangerang Selatan
Workshop Manager: Yusman – firstname.lastname@example.org
The technology workshop at the British School Jakarta follows standard safety measures of a typical school workshop, which are:
I would firstly like to thank my parents, who have consistently been supportive of my love for the sciences and have been pillars of advice and encouragement throughout my research process.
I would also like to thank my physics teacher, Dr. Jeremy Spackman, for introducing me to the Google Science Fair and furthering my love of physics in the two years he has taught me.
Last but not least, I would like to thank the technology department at the British School Jakarta, for without whom I would be still trying to figure out circuitry and CNC routers today.
The assembly of circuitry and all manufacturing processes were conducted in the technology workshop of the British School Jakarta, under adult supervision. The following equipment were available to me and used:
All parts and electrical appliances used in my final product and the testing of my mechanisms were purchased on my own. Testing was conducted out of school under parental supervision.