Week of 3/19/2018 - Progress Report 14

Tamara Jovanovic & Ezequiel Partida
Apr 17, 2018
7 min read

Tamara Jovanovic

This week’s progress was focused on the three following things:

More stability calibration,
Verifying last week’s progress to establish a glucose profile, and
Taking new glucose data.

Last week, we analyzed DI water and 100 mg/dL of glucose at different sensitivity and average waveform settings on the spectrometer. After achieving the maximum power through distilled water, which was about 200 nW, the absorption spectra looked really smooth and nice. Then, all the data was analyzed, and we determined that the best setting for analysis for our project is:

In the zoomed in picture below, it can be seen that the analysis also showed that we have only 0.009 maximum dB difference from the mean, which is really small error margin, indicating that this could possibly be the tolerance margin for our non-invasive system. Since the points on the left were taken at the maximum and minimum, this also shows that any other point on the absorption spectra indicates that there is less than 0.009 dB error on them, and with that even more stability for the system everywhere. The spectra on the right indicates the average spectra of the 100 mg/dL solution and the points on it are the minimum and maximum points on the spectrum.

This analysis could be really crucial for our future glucose analysis. That is why this week, I decided to confirm this data with distilled water, as well as take data of glucose. However, instead of using old glucose concentrations, we decided that we don’t need that big of a difference. The NIR non-invasive glucose project could only produce one point per measurement, and our system can do thousands. So, since we have all this sensitivity, we decided to take glucose measurements closer to those that are actually more realistic for human beings. We still want to re-test our benchmark 100 mg/dL. Since the amount of glucose detected in human blood is between 70 mg/dL and 140mg/dL, we decided to test those two values. Also, just to make sure that we can still get data and analyze it on both ends of the spectra, we also picked two extreme max and extreme minimum values, just to prove our testing. That being said, we also tested 10 mg/dL and 300 mg/dL distilled in DI water.

However, before all of these measurements were taken, and after talking to our project advisor about bettering our system, I decided to take some more steps toward improving even more the stability of our system. That started with printing another new cuvette stand. The old one was good but from use, the side walls started wearing out and moving, since the material in which it is being 3D printed is plastic. Because of the sensitivity of the optical system, even a minor move could alter the measurements and data that needed to be taken. So, I re-printed another cuvette stand, this time ensuring that there will be no movement in it by making the hole for the cuvette surrounded by a really thick wall. The SolidWorks design and actual printed version are shown in the pictures below.

After the new cuvette stand was implemented, measurements were taken. This time around though, I made sure to eliminate some outside factors that were possibly affecting our system. For example, I turned the light off in the optics lab every time measurements were taken. I used a box over our system regardless, and since it has big holes on it for the fiber cable, the light turned off was a good change to make sure no light was going through. Nobody else was working on the optical table, which ensured that the system wasn’t being unnecessarily moved. As per usual, 7 traces of each solution, including DI water were taken at the High 1 sensitivity setting, 2nm resolution and 15 average waveforms. Since we determined that that was the best setting for our system, there was no need for me to take the same data at all the settings available to us, as done in the previous weeks. First, I needed to make sure that I am getting a sufficient amount of power through the Quartz cuvette with DI water in it. Since last week I was getting around 200 nW, that was my goal. It was also to replicate the same absorption spectra. However, since I made changes to the cuvette stand and the reapplication of it to the optical system included some movement of the system, I first needed to work on getting all that power back through DI water. After some time, I was actually able to get almost double the power I got last week: about 390 nW, as shown in the picture below.

This gave me the all ripple absorption spectra, as expected. That spectra is shown below.

I now know that the way to get rid of all the ripples is adjusting the know of the rotational stage, meaning adjusting the cuvette in the cuvette stand, and not touching the collimators. After some turning, the cuvette was a bit angled and my absorption spectra was as smooth as last week. Success!! However, by turning the cuvette at an angle to maximize the absorption spectra, I lost some power, and while tested, I ended up having about 220 nW of power, as shown in the picture below. Also shown is the smooth curve of the absorption spectra at the right angle, looking exactly as good as it did last week.

Once I found that perfect angle to maximize the absorption spectra, I bolted the knob down to make sure that I avoid all unwanted turns and movement to the system.

The cuvette wasn’t taken out of the cuvette stand for the entire process. This also ensured stability. I used the pipette to put substances in and out. I started with water and increased gradually with glucose concentration. According to dr. Li, it is okay to start with the smallest amount of glucose and go up, since that won’t affect the concentrations too much if it is increasing exponentially. The overview of the substances tested is shown in the table below:

All the data looks good but is to be analyzed in Matlab.

For next week’s progress, I will:

Keep taking glucose data,
Make sure that while taking data all the stability steps are taken,
Do more research about absorption spectroscopy
Make sure all the trends are the same, giving us an appropriate glucose profile

Ezequiel Partida

Topics covered: Glucose Spectrum Data Follow Up

Materials used: Matlab, Different Solution Glucose Data

As far as analytical methods go for glucose detection, no progress has been made in recent weeks due to the instability of our optical system and lack of a defined and reproducible glucose spectrum. That is why this week, our focus was solely to continue studying the stability of our system and obtaining consistent glucose spectra by:

1. Reducing the maximum deviation from the mean for distilled water spectra

2. Capturing mean distilled water spectra for a 20 dBm range of interest

3. Capturing subsequent glucose spectra in the same range and settings as water and…

4. Obtaining Glucose Spectrum for different concentrations by subtraction and analyzing them.

1&2. Distilled Water Spectrum and Standard Deviation

Last week, we were able to capture a smooth and consistent spectrum for distilled water by alignment our system, as shown below.

This week, Tamara followed a similar method and was able to get over 300 nW of output for distilled water, so after 7 trials of spectra, the following results were gathered.

As it can be seen, there was an anomaly in the trials captured, so that one trace was later removed from analysis, to obtain a more consistent and accurate representation of our stability for the week.

After removing Trace C, for a wavelength range of around 17 nm and for a 20 dBm range, we were getting around a maximum 0.011 dBm difference in values, which is similar to last week’s 0.01 dBm difference. At this point, it was wise to continue with the data collection process to take spectra of glucose solutions at 10, 70, 100, 140, and 300 mg/dL.

3&4. Glucose Solution Spectra and Subtraction

Last Week’s Spectrum For Reference:

As it can be seen, this week’s glucose data shows a different trend from the preliminary data we took last week. However, this is not entirely bad, since on second inspection, last week’s data shows the glucose spectrum to be higher power than the water spectrum, which does not make sense. This week, all of the glucose spectra we gathered, for all concentrations, were lower power. This makes more sense since a water solution with water should intuitively show more absorbance. The only “bad” aspect of the glucose spectra we calculated is that as seen above, there seems to be a positive linear trend as opposed to a more constant subtraction value across the entire spectrum. Before analyzing this data however, I thought it was useful to create a table and track the points where the subtraction spectrum intersected the average spectra for each respective glucose concentration. The results I observed are shown below.

The table above, in my opinion, shows very promising results. First of all, on the lower intersect points, there is an obvious positive linear trend in the values. That is, as concentration in glucose increases, for the point where the subtraction and average spectra intersect, the subtraction value increases as does the x-intercept point. Additionally, disregarding the order of the 100 mg and 140 mg glucose spectra, for the upper intersect point, there is an obvious negative linear trend in the x-intercept values and an obvious positive linear trend in the subtraction values.

To me, this seems promising and it also makes sense. Earlier this year, Tamara proved that for voltage responses, there is a linear trend in a single wavelength point for glucose concentration. This week, I was able to observe linear trends as well for 2 points (Lower and Upper intercepts in a subtraction spectra scale). However, these 2 points point to a linear trend for the overall spectrum, across several wavelengths, not just 1. Therefore, if the data this week was stable and accurate, we can conclude that we are able to observe easy-to-interpret linear trends in the absorption spectrum of glucose. One more thing that makes this data significant is the fact that the glucose concentrations we used this week are realistic for diabetic cases (70-140 mg/dL). Therefore, we are not reliant on unrealistic cases or faulty data.

Plans for Next Week: