Team:Grenoble-Alpes/Engineering
Detection
It was at Amont à Aussoi that we had met that night to sleep by the fireside. Credits: Estelle Vincent
How to capture fluorescence ?
The first step was to look into existing ways of capturing fluorescence. Photodiodes are commonly used since they convert light into an electrical current, which is produced when photons are absorbed in the photodiode. As surprising as it can be, smartphones are now widely used in medical diagnosis as smartphones’ cameras have the same accuracy as any photodiodes. Plus, as smartphones are part of our daily lives, not only they are an effective solution but a very practical one as well, especially in remote places like Africa. The theory behind the implementation of this technique is as follows. A light filtered by an excitation filter is shed on the object which is positioned perpendicularly to the optical axis. A fluorescent light is back emitted by the object and then filtered by an emission filter. This light is ultimately collected using an optical zoom sticked to the smartphone’s camera.
Filters
The fluorophore wavelength has to be known as precisely as possible so that the appropriate filters can then be acquired. SpectraViewer is a useful tool to simulate the spectral responsivity. Here, the spectral responsivity of the Red Fluorescent Protein (RFP) is simulated.
RFP has a maximum excitation at 555 nm and the maximal emission wavelength is at 584 nm. Therefore, our goal is to excite the fluorophore using a filter centered as close as possible to 558 nm without however having the excitation spectrum overlap the emission spectrum. To achieve this, a bandpass filter has been used for the excitation spectrum, centered on 546 nm. As for the emission filter, it consists of a low pass filter cutting high frequencies. That way the two spectra are clearly distinct, allowing a high precision measurement.
Emission light
For this parameter two options have been considered, either a white lamp or LED. The LED had the advantage of having a very narrow bandwidth and therefore did not require any filter. However, LEDs are low power light sources, which is problematic because a lower power implied a lesser precision in the measurement. For this reason, using a more powerful white lamp combined with the appropriate filter has been prefered.
Optical devices
First and foremost, it is important to note that the excitation light is positioned perpendicularly to the smartphone, as this reduces light scattering. Our optical system is organized as follows: the lamp to which we adjoin the excitation filter, followed by the object and ultimately the adjoined emission filter, lens and smartphone camera.
It goes without saying that since the goal is to detect emitted light, it is of the utmost importance to work in an environment as dark as possible so as to avoid any straylight. Ultimately, as the smartphone’s camera is a little too close from the sample, a lense has been added prior to the camera. This provides us with two big advantages: the image occupies the sensor of the camera, that way more emitted light can be detected, and also the image is sharper, thus taking full advantage of the camera’s quality.
Light intensity
Once a picture has been taken with a smartphone, it needs to be analyzed to provide a numerical value that will help to quantify the amount of fluorescence present. Fluorescence is a value with no unit, it has no reference value which is the reason why a blank must be done serving as the reference value. Therefore, there is no universal way to quantify it and each device could give different values to what is being measured. The goal here is to be able to derive a number from the picture which could inform us on the brightness of the picture, which could in turn provide us information on the amount of fluorescence.
To do so, the picture is analysed using a program on Matlab. The picture is first set to black and white, on a classic 256 gray scale, where 0 is for a black pixel and 255 for the white ones. Then a sum of all the values is done, thus returning a single value. The higher this value is the better, because it would mean that there is more fluorescence overall. As the picture is taken in a dark environment, if a picture appears bright enough it can be due to the fluorescence. It is important to note that for each picture the frame is the same, the tubes are placed at the same place implying that the surface captured each time is the same. We can therefore evaluate the amount of fluorescence by applying the calculation to the entire picture and compare the values obtained. The numerical value obtained can then be compared to one obtained using a measurement instrument, such as a spectrophotometer, and conclude if the method and the overall system are reliable.
TESTS REALISED
The goal of the test is to verify that the detection of fluorescence is occurring as expected and determine if there is a certain threshold under which fluorescence cannot be detected anymore.
Protocol and implementation
- 6 test tubes are prepared, each having the same volume and optical density (OD) so as their fluorescence can be compared.
- Fluorescent bacteria are used as well as non-fluorescent. In fact, the OD of each tube has to be the same since the less there are fluorescent bacteria, the smaller the value of OD is, that’s why it is compensated by adding nonfluorescent bacteria. LB is used to compensate the volume, so each tube has the same volume, otherwise a comparison wouldn’t be possible.
- The tube’s composition is as follows.
| 1 | 2 | 3 | 4 | 5 | 6 | |
| Targeted plasmid | 0.00 | 0.02 | 0.04 | 0.06 | 0.08 | 0.1 |
| Non fluorescent bacteria JM109 | 0.50 | 0.419 | 0.338 | 0.257 | 0.176 | 0.094 |
| LB | 0 | 0.061 | 0.122 | 0.183 | 0.244 | 0.305 |
- A picture of each tube is then taken individually, with the 3D printed kit. Notice that a cover is put above the kit to create a dark environment.
- The pictures are then analysed using the Matlab program, which informs us of the intensity of fluorescence detected on the picture.
- So as to verify that our device is working properly, measurements are made at the same time with a spectrophotometer, allowing for a comparison with the result obtained with the smartphone.
Results
Before making any other comment, it is important to emphasize on the fact that since the same OD were obtained in each tube, it is reasonable to compare the results.
| Tubes | LB | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|---|
| Optical Density | 0 | 0.375 | 0.270 | 0.265 | 0.270 | 0.260 | 0.290 |
What is particularly interesting in these results is that we have the possibility to compare them with those of a calibrated spectrophotometer. As shown in the graphs below, both of the measurement devices have a linear variation between the quantity of fluorescent bacteria and the level of fluorescence measured. In fact, we use a smartphone with a camera that contains a light sensor proportional to the light intensity. That being said, it enables us to compare these two devices. It can be seen on the graphs that in both cases, the fainter the fluorescence, the greater the incertitude associated. At this point, there is no threshold, because even though the incertitudes are greater for smaller values, they are of lower magnitude compared to incertitudes of greater values. Thus we can take into account every fluorescence measured.
Discussion
Once the fluorescence detector built appears as an accurate device, it is useful to compare its returned values with the ones returned by the spectrophotometer in a more precise manner. Measurements of fluorescence of the six tubes taken by the spectrophotometer is plot in function of the measurements taken with the smartphone. As it can be seen, the values obtained using the kit vary in the same way as those obtained with the spectrophotometer. As it is safe to assume that the spectrophotometer is returning sensible data, the experimental data can also be expected to be correct.
What would be interesting here is to calibrate our device, in order to be able to adjust any other smartphones to this kit. The calibration of a device stays an important part of its fabrication and it is important to consider whether the measurement is reproducible or not. Even if that is not our case, we wanted to find a way to calibrate every SnapLab devices. For that purpose, a small sheet of red paper can be delivered with the kit, thus allowing each user to properly calibrate the kit independently of the smartphone they are using. The first step would be to take a picture of this red colored paper with the smartphone used for the analysis. The application could then be able to adjust the sum of the pixels’ intensities to the one given by the device we built with our smartphone.
It is worth mentioning that results can only be considered as correct if they are located within the errorbars. The uncertainties have been evaluated as follows: three pictures of the same test tube were taken, over the entire tube tests, that way we can determine if the uncertainty depends on the intensity of the fluorescence or not, just as it would for a spectrophotometer. We then evaluated the standard deviation, and depending on the results, a specific errorbar is given or a single errorbar is considered for the entire spectrum by taking the mean of all the standard deviations. Same work was made for the spectrophotometer in order to compare the results.
| 1 | Non fluorescent tube | Fluorescent tube | |
| Human reproducibility | 91.63% | 90.11% | 95.76% |
| Spectrophotometer reproducibility | 91.64% | 90.80% | 97.79% |
| SnapLab reproducibility | 99.54% | 99.84% | 99.97% |
This allows us to comment on the precision of the smartphone’s camera. Concerning the level of fluorescence in each tubes, the more fluorescent it is, the more precise the measurement will be. However, taking into account a minimum threshold is not necessary because measurements stay accurate even if there are less precise. Concerning the values of uncertainties obtained with the smartphone, it appears to be very small. In fact, unlike the uncertainties of the spectrophotometer which are bigger, the experiment made to determine error bars for the smartphone was quite restrictive. Only a few uncertainties were considered, leading to a huge level of reproducibility. To improve this value, other tests could be done considering other sources of uncertainties. It would lower the level of reproducibility but not drastically.
Finally, as it can be seen, it can be assumed that the telephone inside the kit is a real measurement tool.
