*4.2. Experimental Results*

The simulation results showed that the developed sensor is able to distinguish between different levels of blood glucose concentration. The sensor was tightly attached to the test sample, and the simulation showed the best result at a frequency close to 1 GHz.

An experimental study showed that the maximum difference in amplitudes is in the frequency range 1.45–1.55 GHz (Figure 13). This bias is due to the inaccuracy of the sensor manufacturing compared to simulation (1.07 GHz). Considering this range, it can be seen that the amplitudes at different concentrations vary.

**Figure 13.** Dependence of the reflected signal on frequency for a near-field sensor in the frequency range 0–5 GHz.

A detailed examination (Figure 14) showed that the reflected signals are lined up in the correct sequence, but the difference between the values of 7 and 9 mmol/L is extremely small. This result is associated with the non-linear behavior of the dielectric permittivity in the frequency range of 1–2 GHz (Figure 3).

**Figure 14.** Dependence of the reflected signal on frequency for a near-field sensor in the frequency range 1.3–1.57 GHz.

The rest of the concentrations are arranged in order, which indicates the correctness of this approach to the study of the concentration of glucose in the blood. Table 4 shows the average values of the amplitude of the reflected signal from the sensor for different glucose concentrations at a frequency of 1.53 GHz. Based on the results of the 10 conducted experiments, a confidence interval was constructed with errors for each individual glucose concentration value.

**Table 4.** The results of measuring the amplitude of the reflected signal by the proposed new sensor for different glucose concentrations.


As can be seen from the table, the values of the amplitude of the reflected signal with a change in concentration differ by 0.15–0.4 dB; on average, the amplitude is 0.27 dB. The obtained sensitivity exceeded the simulation results (0.1 dB).

Based on the obtained results, the dependence of the reflected signal (reflection coefficient S11) on various glucose concentrations in saline solution is plotted (Figure 15). The crosses mark the data obtained experimentally at a frequency of 1.53 GHz. A regression line is drawn along these experimental data, and an error interval was built for them taking into account the data presented in Table 4. It can be seen that at the minimum glucose concentration, the obtained values are aligned according to the regression curve. Furthermore, the concentration of 10 mmol/L matches the regression curve. In the case of concentrations of 7 and 9 mmol/L, the values differ from the expected ones, so the maximum difference is seen at a sugar level of 9 mmol/L. This deviation is associated with the non-linear behavior of the dielectric constant value at low frequencies (Figure 3).

**Figure 15.** The dependence of the reflected signal on the concentration of glucose in saline at a frequency of 1.53 GHz.

Below is a comparison of the data obtained using the developed sensor with other developments (Table 5). To compare the results of this study with others, we selected the results where both the reflected signal and the transmitted signal were used as the measured parameters.


**Table 5.** Comparison of the developed sensor with other modern sensors for determining glucose levels.

In [18], a microstrip glucose sensor measured the reflected signal and the input impedance. The authors managed to achieve a sensitivity of 0.18 dB/(mg/dL) at a frequency of 1.4–1.9 GHz. In [19], a method based on the change in the reflection coefficient between the dielectric waveguide resonator and the measured liquid was used. The proposed sensor allows detecting changes of 0.5 mg/mL at a sensor sensitivity of 0.003 dB/(mg/mL).

Among the presented sensors from Table 5, the most sensitive is the sensor of [21]. It is based on the measurements of the transmitted signal with resonant frequencies in the range of 50–70 GHz. This method is based on the shift of the resonant frequency. The sensitivity of the sensor was 0.8–1 dB/(mg/mL) in an experimental study and 7.7 dB/(mg/mL) in simulation. Comparing the obtained data with the literature data, we can talk about the high sensitivity of the presented sensor in comparison with others. The proposed near-field sensor for non-invasive glucose monitoring surpassed all sensors in Table 5 at its sensitivity, which was 1.7–3.4 dB/(mg/mL).
