Design:

The first step in starting to fabricate a microchannel is the design. We used drafting software to draw the design of the microchannel, taking into account the flow path of the fluid and the required functions.


where d, D1, D2, D3 are the cell flow concentration widths, v1, v2 and v3 are the fluid inlet flow velocities, and 𑄺a, 𑄺1, 𑄺2 and 𑄺3 are the fluid densities of the injected fluid at the outlet and at the inlet microfluidic channel.


Based on this formula, we designed a trident-shaped microchannel, with the main flow channel at 150 microns and the other two channels at 300 microns, constricted to 20 µm.

Making Shells:

The design of the trident microchannel is converted into a digital model by slicing software for printing shells through a light-curing 3D printer. The printed shells are solidified by UV-irradiation.

Shell casting:

The mixture of ethylbenzene and PMS (10:1) is degassed by a vacuum machine and injected into a microchannel shell.

Solidifying the microchannel:

The casting shell of the microchannel is baked at 85 C for 4 hours for solidification. The solidified microchannel is carefully removed from the shell.

Fixing the microchannel:

The removed microchannel was attached to a carrier slide and exposed to ozone for 60 min to ensure complete crosslinking between the microchannel and carrier slide.

Ozone Processing:

Combine the microchannel with the carrier slide and expose it to ozone for 60 minutes to ensure that the microfluidic channel is fully bonded to the carrier slide.

(All experiments are conducted in a dimly-lit room to minimize errors and facilitate highly accurate data collection.)

Parameter curves of opto-electronic diodes

The first step of building is to test the limitation of photodiodes, and to establish a reference of the photodiode's voltage output. To test the limitations of photodiode, we decreased the light intensity as below:

• Polarizers are employed to regulate the luminosity of the emitted laser light
• A piezoelectric tank is employed to enhance the energy dispersion of the transmitting light.
• Multiple acrylic sheets are applied to decrease light intensity by refracting and reflecting.

After setup, the decrease of light intensity was detected by optical power meter, showing that our setup could decrease the laser's intensity from a minimum of 5mW to the lowest detectable 10nW.

Next, we replaced the optical power meter with a photodiode, and recorded the measurement displayed on the triple meter. In theory, the voltage generated by the photodiode should increase with the intensity of light. To analyze the data, we selected three experiments closely aligned with the theoretical data out of the ten conducted and calculated their average.

Amplified Signal

The voltage detection module was integrated with the photodiode, and Arduino was used to measure the voltage value. It was observed that surpassing a specific threshold was necessary in order to register a value that could be detected by Arduino. However, due to the limited brightness of the fluorescent material, it was unable to reach the required threshold. Consequently, amplification of the voltage generated by the photodiode was needed.

Initially, we used the commonly available LM358 amplifier. However, our attempt to amplify the photodiode using the inverting amplifier circuit did not yield the expected outcome as it failed to amplify the photodiode in the way we had anticipated.

▲ No discernible difference without (left) and with amplifier circuit (right)

After searching for related papers, we found that this result is due to LM3588 being a general-purpose amplifier, so its amplification precision is not enough to amplify small signals. Therefore, we changed to an instrumentation amplifier, AD620, and referred to the circuit diagram below for trial use. This time it successfully amplified the small signal generated by the photodiode.

▲ Voltage before amplification was 0.0008V, after amplification, 0.392V, a 490-fold amplification

Before amplification, the voltage was 0.0008V and after amplification, it increased to 0.392V, resulting in a 490-fold amplification. Accordingly, we selected the AD620 amplifier module for subsequent detection.

Measurement of Fluorescent protein

After confirming the photodiode's parametric curve and the amplification circuit, we replaced the laser with stimulated mGL solution as a light source for measurement.

We placed the mGL solution in a quartz tube, and stimulated it by a 488 nm laser. The laser will pass through the visible window and stimulate the mGL in the quartz tube.


To detect the light emitted by th stimulated mGL, we used a bandpass filter (500-525 nm) and adhered the photodiode next to the filter. We also printed a holder for filters and photodiodes by 3D printer. To prevent light leakage, we covered the outer surface with black electrical tape.


Next, we obtained crude mGL protein solution from the wet lab, and serially diluted the protein solution into five concentrations: 1.83*10^-7 mg/ml, 9.14*10^-8 mg/ml, 1.83*10^-9 mg/ml, 9.14*10^-10 mg/ml, and 1.83*10^-11 mg/ml. The measurement result after laser stimulation shows that the voltage difference decreases as the protein concentration decreases.

Mimicking CTC measurement

We then tried to test if we could measure the light emitted from the mGL-labeled CTCs. Since we were not allowed to collect clinical samples, we decided to mimic the mGL-labled CTCs. On average, each CTC carries 5.00E+5 FRα molecules (Hongyun Zheng, 2021). We assumed that all the FRα are recognized by mGL. That is to say, we could mimic one CTC if we put 5.00E+5 mGL molecules in the optical path. Accordingly, we estimated the concentration of mGL should be 4.15E-18~4.15E-17.



After calculating mGL concentration corresponding to different numbers of labeled CTCs, we carried out relevant examinations.



Unfortunately, the result indicated that the voltage difference is not proportional to the number of cells, and no distinguishable pattern can be discerned between the sets of data. It is possible that the concentration is undetectable, or there may be interference caused by noise.

To examine whether amplifier noise affects the result, we measured the voltage difference without the amplifier. The result still shows that the measurement value is not proportional to the number of cells. Therefore, we concluded that the photodiode we used may not be sensitive enough.