A Novel Low-Cost Capacitance Sensor Solution for Real-Time Bubble Monitoring in Medical Infusion Devices

Abstract

In the present day, IoT technology is widely applied in the field of medical devices to facilitate real-time monitoring and management by medical staff, thereby better-ensuring patient safety. In IoT intravenous infusion monitoring sensors, it is particularly important to ensure that air bubbles are not infused into the patient’s body. The most common method for bubble detection during intravenous infusions is the use of infrared or laser sensors, which can usually meet design requirements at a relatively low cost. Another method is the use of ultrasonic detection of bubbles, which achieves high accuracy but has not been widely promoted in the market due to higher costs. This proposed work introduces a new type of sensor that detects bubbles by monitoring changes in capacitance between two electrodes installed at the surface of the infusion pipe. If this sensor is deployed on the ESP32 platform, which is widely used in embedded IoT devices, it can achieve 35 μL bubble detection precision with an average power consumption of 5.18 mW and a mass production cost of $0.022. Although the precision of this sensor is significantly lower than the low-cost IR bubble sensor, it still satisfies the design requirement of the IV infusion IoT sensor.

1. Introduction

Intravenous therapy (IV therapy) is a medical procedure that injects fluids, drugs, and nutrients straight into a patient’s vein [1,2]. The use of IoT technology to monitor and even take over [3] the intravenous infusion process ensures the safety and accuracy of the treatment [4,5,6]. Thus, it is widely used in hospitals and even in patients’ homes [7,8]. With the proliferation of IoT technology, an increasing number of requirements have emerged [9,10,11,12,13], including the timely detection of air bubbles during intravenous infusions. During the process of infusion, it is often inevitable that air bubbles enter the infusion pipe. These large air bubbles may block blood vessels and increase the burden on the heart [9,14]. Common IV infusion bubble detection methods typically include optical and ultrasonic methods. Both of these methods can achieve good sensitivity, exceeding the design requirements. This article introduces a method for bubble detection by measuring the capacitance between electrodes installed on either side of the infusion pipe and simulated scenarios that may arise for actual use to demonstrate the viability of this method. When this method is applied to the ESP32 IoT platform, it satisfies the design target and significantly reduces costs compared to the optical and ultrasonic methods. The following points summarize the uniqueness and innovation of this proposed work:

• With the touch io feature of the ESP32 chip, the cost of the sensor is extremely low compared to common infrared bubble sensors while maintaining enough accuracy.
• This bubble detection method can be easily ported to some non-standard infusion pipes at an almost unchanged cost while still meeting the basic design requirements for accuracy.
• The power consumption of this method is lower than that of infrared bubble sensors due to its use of ESP32’s built-in hardware layer function to realize.

2. Materials and Methods

2.1. Overview: Importance of Bubble Detection

In medical IoT applications, the advantage of IoT is the real-time monitoring of various patient conditions. Bubble detection in infusion monitoring is particularly important [15,16]. Studies show that, in intravenous infusion, large-sized air bubbles mistakenly infused into a patient’s body might block blood vessels, increasing the burden on the heart. However, air bubbles smaller than 50 μL will not cause significant adverse effects [17]. As shown in

Table 1. Products’ features comparison.

Products	Features
	Flow Rate and Volume Monitoring	Alarm Warning	Air-in Line Detection	WLAN	Pressure Anomaly Detection	Automatic Infusion
Rainbow Care IV Drip Stand and Unigloves Infusion Set	No	No	No	No	No	No
Shift Labs DripAssist	Yes	Yes	No	No	No	No
B. Braun Easypump II ST/LT	No	No	No	No	No	Yes
Alaris GH Plus Guardrails Syringe Pump	Yes	Yes	No	No	Yes	Yes
Agilia SP MC WiFi	Yes	Yes	No	Yes	Yes	Yes
Alaris GP Plus Volumetric Pump with Guardrails	Yes	Yes	Yes	No	Yes	Yes
Agilia VP MC WiFi	Yes	Yes	No	Yes	Yes	Yes
Moog CURLIN 6000	Yes	Yes	Yes	No	Yes	Yes
Moog CURLIN PainSmart IOD	Yes	Yes	Yes	No	Yes	Yes

, a significant number of smart infusion devices are already equipped with bubble detection functionality. In medical devices, such sensors tend to use non-contact methods [18,19]. Most IoT-based smart infusion devices currently use infrared [20] or laser [21] sensors to detect bubbles. In our experiment set, we observe if the IR emitter is too close to the infusion pipe, the IR rays emitted will almost entirely reach the receiver along the pipe walls, regardless of whether there is air inside the pipe. As shown in Figure 1, due to interference from the infusion pipe light waveguide, this method often fails to detect some small bubbles less than 10 μL. Another limitation of optical sensors for bubble detection is that a certain design of an optical bubble sensor is only for certain dimensions of the infusion pipe. If the design is directly used on other dimensions of the pipe, it might cause a blind zone. An alternative, more accurate, but costly method is to use ultrasonic sensors for bubble detection [22], although no such smart infusion products are currently available in the market.

2.2. Proposed Use of Capacitive Sensors in IoT Medical Device

A simple cylindrical capacitor consists of two curved plates and a dielectric in between. Figure 2 shows a cylindrical capacitor with two different dielectrics. The capacitance at the ends of the plates follows the given formula [23].

$$dC=\frac{{\epsilon }_0{\epsilon }_r{\epsilon }_{rx}la}{2{\epsilon }_{rx}\left(a-b\right)+2b{\epsilon }_{r}cosϴ}$$

(1)

where ε_r is the relative dielectric constant of a non-metal pipe and ε_rx is the relative dielectric constant of the material contained in a non-metal pipe. When an air bubble is present between the electrodes, the dielectric inside the pipe changes, and the capacitance between both plates also changes as given formula [24].

$$\Delta C=-\frac{V}{{3a}^2}\frac{{\epsilon }_{rx}-1}{{\epsilon }_{rx}+2}{\epsilon }_0$$

(2)

where V is the volume of the air bubble.

Capacitive sensors are already used in IV infusion to monitor fluid [23,24,25,26,27] or combined with PDMS membrane technology to monitor blood pressure, oxygen saturation, and heart rate [28,29,30,31]. Currently, such methods are used for flow rate and patient indication monitoring, while this study applies the method for bubble detection. This study proposes a method to directly use ESP32 to read sensor readings, eliminating the need for additional peripheral circuits, simplifying the design, and reducing costs.

2.3. Advantages of Capacitive Sensors for Infusion Bubble Monitoring

Compared to traditional optical methods, capacitive sensors can achieve similar accuracy at lower cost and power consumption. This is especially true for the widely used ESP32 platform in various IoT devices [32,33]. The ESP32 chip features touch IO ports dedicated to measuring capacitive changes, typically used in capacitive touch screens [34,35]. With the appropriate configuration of the touch IO ports [36], detecting bubbles in the infusion pipe is possible using just two copper foil electrodes, making it cost-effective.

2.4. Methodology of Studies

The sensor’s structure is very simple, as shown in Figure 2. It only requires two copper foil electrodes with connecting wires shielded by a dielectric material. One wire connects to the ESP32 ground, and the other to the ESP32’s touch IO port. The capacitance changes between two electrodes cause variations in the microcontroller touch readings. By analyzing the touch readings from the microcontroller, it is possible to determine whether large air bubbles are present in the tube. The formula suggests that the length of the copper foil is not highly related to the detectable volume of the bubbles, but the width of the foil does affect sensitivity. Considering installation difficulties and potential errors, a length of 15 mm is chosen for the following experiment. A wider foil is better if space permits, so a higher width will be chosen for a wider pipe. With the ESP32 touch IO configuration, while touch IO ports connect to the sensor, the measuring frequency is between 34.5 and 40 kHz (with an additional 16 pF scope probe capacitance), so all measurements are conducted under a frequency of 40 kHz. All dimensions of infusion pipes are from LEIRONG Technology; copper foil is from SUNWAY Precise Metal, and Polyimide tape is from DuPont. All studies were conducted at temperatures between 21 and 28 °C. For those measurements using an LCR meter, the copper foils were connected with 10–11 cm of 18 AWG copper wires, and a TH2818 LCR bridge was used to measure. For the part tested with the ESP32 development board, an additional 30 cm Dupont jumper wires were used to make it easier to connect the sensor to the development board. A certain volume of air bubble is injected into the infusion pipe by a high-precision infusion pump. This infusion pump can adjust the flow rate in steps of 0.12 μL/min, thus providing the sufficient accuracy required for this experiment.

3. Design and Development

3.1. Effectiveness for Different Types of Infusions

Intravenous infusions, a common clinical treatment, involve various types of liquids like isotonic, hypotonic, hypertonic, and colloidal solutions [37]. When the pipe is filled with liquid, the capacitance will increase with the increase of the inner dielectric constant. Different solutions have different conductivities and permittivity, which might affect the measurement’s validity. For isotonic, hypotonic, and hypertonic infusions, the medication concentration is typically low enough not to significantly affect the liquid’s electrical properties.

Table 2. Guidelines for intravenous medications.

Drug Name	Std Concentration	Std Dilution
Abciximab	36 μg/mL	9 mg/250 mL
Amiodarone	1.8 mg/mL	450 mg/250 mL
Bumetanide	0.04 mg/mL	10 mg/250 mL
Cisatracurium	0.8 mg/mL	200 mg/250 mL
Dexmedetomidine (Precedex)	4 μg/mL	200 μg/50 mL
Diltiazem (Cardizem)	1 mg/mL	100 mg/100 mL
Dobutamine (Dobutrex)	2 mg/mL	500 mg/250 mL
Dopamine	1.6 mg/mL	400 mg/250 mL
Epinephrine	0.02 mg/mL	5 mg/250 mL
Esmolol	10 mg/mL	2500 mg/250 mL
Fentanyl	10 μg/mL	1000 μg/100 mL
Heparin	100 units/mL	2500 units/250 mL
Insulin (regular)	1 unit/mL	100 units/100 mL
Lidocaine	4 mg/mL	1000 mg/250 mL
Magnesium	20 mg/mL	1000 mg/50 mL
Morphine	1 mg/mL	50 mg/50 mL
Nicardipine	0.1 mg/mL	25 mg/250 mL
Potassium Chloride	0.1 mEq/mL	10 mEq/100 mL
Propofol (Diprivan)	10 mg/mL	1000 mg/100 mL
Vasopressin	0.4 unit/mL	20 units/50 mL

lists some common infusion types and their drug concentrations obtained through medical centers and organizations (UofL, University of Illinois Medical Center, etc.) [38,39,40,41]. Colloidal fluids like plasma and albumin are often diluted at high concentrations or not at all [42], making them difficult to uniformly assess.

Table 3. Measurement of different kinds of content in pipe.

Content	Cp (at 40 kHz)
Air	1.35 pF
Isotonic solution (0.9% saline solution)	1.80 pF
Hypotonic solution (0.45% saline solution)	1.82 pF
Hypertonic solution (3% saline solution)	1.71 pF
Isotonic solution (5% glucose solution)	2.29 pF
Hypertonic solution (10% glucose solution)	2.07 pF

shows the Cp measurements of different solutions and air in a medical silicone pipe (outer diameter 6.4 mm, inner diameter 3.2 mm) with electrodes (15 mm by 9 mm).

The results cover most therapeutic isotonic, hypotonic, and hypertonic fluids and show significant differences from air.

3.2. Effectiveness for Different Infusion Pipe Sizes

Table 4. Guidelines for intravenous medications about dosing rate.

Drug Name	Dosing Rate	Flow Rate (For 65 kg Adult)
Abciximab	0.125 μg/kg/min	0.226 mL/min
Amiodarone	1 mg/min	0.556 mL/min
Bumetanide	0.25–0.5 mg/h	6.25–12.5 mL/h
Cisatracurium	0.15–0.2 mg/kg/h	12.189–16.25 mL/h
Dexmedetomidine (Precedex)	0.2–1.4 μg/kg/h	3.25–22.75 mL/h
Diltiazem (Cardizem)	5–15 mg/h	5–15 mL/h
Dobutamine (Dobutrex)	2.5–20 μg/kg/min	0.081–0.65 mL/min
Dopamine	2–20 μg/kg/min	0.081–0.812 mL/min
Epinephrine	0.05–0.45 μg/kg/min	0.163–1.463 mL/min
Esmolol	50–300 μg/kg/min	0.325–1.95 mL/min
Fentanyl	25–150 μg/h	2.5–15 mL/h
Heparin	13 units/kg/h	8.45 mL/h
Insulin (regular)	Based upon blood glucose level–follow the protocol	-
Isoproterenol	0.5–20 μg/min	0.031–1.25 mL/min
Lidocaine	1–4 mg/min	0.25–1 mL/min
Magnesium	1000 mg/h	50 mL/h
Morphine	1–10 mg/h	1–10 mL/h
Nicardipine	5 mg/h	50 mL/h
Potassium Chloride	$\le$10 mEq/h	$\le$100 mL/h
Propofol (Diprivan)	5–75 μg/kg/min	0.033–0.488 mL/min
Vasopressin	0.04 units/min	0.1 mL/min

lists common therapeutic flow rates [38,39,40,41], indicating that pipes with an inner diameter as small as 3 mm can meet flow requirements. However, there are views that 3 mm pipes may not suffice for some emergency treatments that require rapid fluid infusion [43], and various sizes of medical infusion pipes are available on the market.

Table 5. Measurement of different dimensions of pipe.

Dimensions	ΔCp (at 40 kHz)	Electrode Dimensions
6.4 mm × 3.2 mm	0.45 pF	15 mm × 9 mm
9.6 mm × 6.4 mm	0.34 pF	15 mm × 12 mm
8 mm × 4.8 mm	0.38 pF	15 mm × 11 mm
7.2 mm × 4 mm	0.48 pF	15 mm × 10 mm
6 mm × 4 mm	0.37 pF	15 mm × 8.5 mm
5.6 mm × 2.4 mm	0.2 pF	15 mm × 8 mm
5 mm × 3 mm	0.21 pF	15 mm × 7 mm
4.8 mm × 1.6 mm	0.18 pF	15 mm × 7 mm
4.5 mm × 2.5 mm	0.28 pF	15 mm × 6.5 mm
4 mm × 2 mm	0.30 pF	15 mm × 6 mm
3.5 mm × 1.5 mm	0.06 pF	15 mm × 5 mm
3.2 mm × 1 mm	0.09 pF	15 mm × 5 mm

shows the change in Cp measurements for different sizes of medical silicone pipes and dimensions of electrodes.

The capacitance difference is more significant in a thick pipe, whereas in a thin pipe, the changes are not as noticeable but can still be detected by the microcontroller.

3.3. Earth Effect on Capacitance

As shown in Figure 3, due to the higher potential on the plates compared to the Earth, the charge on the plates couples to the Earth, forming an undesired additional capacitance to the space [44]:

$$\mathrm{C}1=2\mathsf{\pi }\mathsf{\xi }0/\mathrm{l}\mathrm{n}(2\mathrm{h}/\mathrm{r})\times l$$

(3)

where h is the distance to the earth, r is the sum of the radius of the infusion pipe and copper thickness, and l is the copper foil length.

Table 6. Measurement of different distances to the Earth.

Height	ΔCp (at 40 kHz)
0	0.78 pF
0.05 m	0.51 pF
0.1 m	0.12 pF
0.15 m	0.06 pF
0.2 m	0.03 pF
0.25 m	0.02 pF
0.5 m	0.02 pF

shows measurements of capacitance change (refer to 1 m above the ground) at different heights above the ground for saline inside a medical silicone pipe (6.4 × 3.2 mm) with electrodes (15 mm by 9 mm).

The results indicate almost no impact at 0.15 m above the ground, which is much lower than the height at which patients receive treatment, suggesting that the ground effect does not significantly impact sensor accuracy.

3.4. Electric-Field Screening

As shown in Figure 4, when another conductor with a lower potential and proximity to the electrode appears, the plate charges the conductor, altering the capacitance reading. Also, noise from the conductor may couple into the sensor. Therefore, electric-field screening of the electrodes is necessary [45].

Polyimide tape, commonly used in industry for insulation and heat resistance [46], can prevent other conductors from getting too close to sensitive points in the circuit.

Table 7. Measurement of different thickness polyimide tape screening effectiveness.

Thickness	ΔCs (at 40 kHz)
0 (touch on single electrode)	4.61 pF
0 (touch on both electrodes)	186.82 pF
0.1 mm (touch on single electrode)	0.13 pF
0.1 mm (touch on both electrodes)	0.10 pF
0.2 mm (touch on single electrode)	0.11 pF
0.2 mm (touch on both electrodes)	0.10 pF

shows the impact of direct hand contact on sensor readings when different thicknesses of polyimide tape are used for screening over the electrodes, with other conditions being the same as in previous tests.

The result shows a 0.2 mm thick polyimide tape is sufficient to prevent direct hand contact from affecting the readings.

3.5. ESP32 Touch IO Port Configuration

Espressif provides detailed documentation for its ESP32 series microcontroller. According to the documentation [36], the reading procedure for the touch IO port is shown in Figure 5. There are two concepts often mentioned when discussing the configuration of the ESP32 touch IO ports: IIR and FSM. The infinite impulse response (IIR) filter is a commonly used digital signal processing filter. When configuring an IIR filter for the ESP32 touch IO ports, it acts similarly to a low-pass filter to reduce noise and fluctuations in the measurement results. A finite state machine (FSM) is an algorithm that switches between different states based on inputs. In the configuration of the ESP323 touch IO ports, the FSM timer is used to control state transitions when the timer reaches a given time. Once measurement begins, the IO port charges until reaching Vrefh, which is the setting high level. Then, discharges are made until the low-level Vrefl is set. This process repeats throughout the measurement time, and the number of high-low cycles is returned at the end of the measurement duration.

The capacitance of the circuit affects the charging and discharging speed, and the speed change will reflect on the measurement result. The ESP32 allows eight different slopes for charging and discharging the touch IO port. However, no difference was observed for each slope configuration in the practical test, so the slope was set to default.

Table 8. 1000 measurements variance of different voltage setting combinations.

	Vrefl	0.5 V	0.6 V	0.7 V	0.8 V
Vrefh
2.4 V		0.000012	0.000003	0.0005	0.000002
2.5 V		0.000005	0.00001	0.000013	0.000001
2.6 V		0.000003	0.000005	0.000051	0.00002
2.7 V		0.000001	0.000002	0.000006	0.000021

shows the normalized variance of 1000 measurements without a filter at different high-low voltage level settings. The results indicate good stability when setting the high level at 2.5 V and the low level at 0.8 V.

After configuring the WiFi, the normalized variance significantly increased. Under the configuration mentioned above, turning on WiFi increased the normalized variance of 1000 readings from 0.000001 to 0.000015. If a 20 ms period IIR filter that touches IO built-in is configured, the normalized variance decreases back to 0.000002. Under the above configuration, setting up a 20 ms period IIR filter results in sensory response and recovery times of approximately 300 ms. Although it is relatively long, since the flow speed of the infusion is generally very slow (less than 4.15 mm/s), a 20 ms period IIR filter will not affect accuracy. From the result above, the touch IO port is configured to default slope, with a high level of 2.5 V, a low level of 0.8 V, and a 20 ms period IIR filter. Under this configuration, measurements were taken for the sensor with and without about 50 μL air bubble during saline infusion, with readings of 51,600 and 52,700. Therefore, the bubble detection threshold is set to 1%, deviate the initial value. The system flowchart is rough, as shown in Figure 6. When infusion starts, the ESP32 reads the touch IO port 10 times to establish a baseline value. If the touch IO port reading deviates from the baseline beyond the threshold, it indicates a bubble in the infusion pipe, triggering an alarm to the IoT terminal. Furthermore, the microcontroller can also stop the infusion to prevent the bubble from entering the body.

4. Result

4.1. Detection of Different Volumes of Air in Various Solutions

A high-precision pump on the market is used to deliver quantified air into a tube to study the sensitivity of bubble detection.

Table 9. Sensitivity for different solutions.

Air Volume	Isotonic Solution (0.9% Saline Solution)	Hypotonic Solution (0.45% Saline Solution)	Hypertonic Solution (3% Saline Solution)	Isotonic Solution (5% Glucose Solution)	Hypertonic Solution (10% Glucose Solution)
10 μL	Failed to detect	Failed to detect	Failed to detect	Failed to detect	Failed to detect
15 μL	Failed to detect	Failed to detect	Failed to detect	Not stable	Not stable
20 μL	Not stable	Not stable	Failed to detect	Not stable	Stably detect
25 μL	Not stable	Not stable	Failed to detect	Stably detect	Stably detect
30 μL	Stably detect	Stably detect	Not stable	Stably detect	Stably detect
35 μL	Stably detect	Stably detect	Stably detect	Stably detect	Stably detect

shows the sensitivity of testing different volumes of air in various solutions inside a 6.4 × 3.2 mm pipe. It is evident that a volume of 35μL of air can be stably detected in all three solutions, so the working range of this sensor is >35 μL. The sensitivity of the MCU is approximately 4 units/fF.

4.2. Comparison with Low-Cost Optical Methods

Figure 7 shows a design for a low-cost optical method. Testing for air bubbles in a 6.4 × 3.2 mm infusion pipe can stably detect a volume of 10 μL of air, obtaining higher accuracy than the capacitive method.

Table 10. Low-cost IR bubble sensor BOM.

Electronic Components	Quantity	Cost
OP999 (3 mm IR photodiode)	2	$2.12 (for a quantity of 3000+, it is $1.79)
LM358P	2	$0.91 (for a quantity of 2500+, it is $0.42)
SFH4550 (3 mm IR LED)	2	$0.85
Passive components (8 resistors, 4 capacitors)	-	$0.03
Materials	Quantity	Cost
Wire	-	-
PLA+ 3D printer filament	-	$0.5 (if ordering a quantity of 3000+ from the factory, it is $0.17)
Subtotal		$4.41 (for a quantity of 3000+, it is $3.26)

and

Table 11. Capacitive bubble sensor BOM.

Electronic Components	Quantity	Cost
-	-	-
Materials	Quantity	Cost
Wire	-	-
Copper foil (15 mm × 9 mm)	2	$0.05 (for a quantity of 3700+, it is $0.017)
Kapton tape (19 mm × 30 mm)	1	$0.006 (for a quantity of 5500+, it is $0.005)
Subtotal	-	$0.056 (for a quantity of 5500+, it is $0.022)

show the low-cost IR bubble sensor BOM and capacitive bubble sensor BOM respectively. The power consumption of this sensor is 11.8 mW. The power consumption of ESP32 without and with configured touch IO is shown in Figure 8. The average power consumption difference is 5.18 mW.

Table 12. Comparison of low-cost IR bubble sensor and capacitive bubble sensor.

Sensor Type	Capacitive Bubble Sensor	Low-Cost IR Bubble Sensor
Sensitivity	35 μL	10 μL
Cost	$0.056 (for a quantity of 5500+, it is $0.022)	$4.41 (for a quantity of 3000+, it is $3.26)
Power Consumption	5.18 mW on average	11.8 mW
Size	Almost the pipe size	30 mm × 10 mm × 15 mm
Weight	<0.8 g (wire excluded)	About 16 g (wire excluded)

below lists the overall comparison of the capacitive bubble sensor and low-cost IR bubble sensor.

4.3. Impact of Larger Electrode Width on Sensitivity

Table 13. Sensitivity for different solutions.

Dimension	ESP32 Threshold	Sensitivity
6.4 mm × 3.2 mm	1%	30 μL
9.6 mm × 6.4 mm	1.8%	50 μL (unable to prevent hand contact interference)
8 mm × 4.8 mm	2%	50 μL
7.2 mm × 4 mm	1.2%	35 μL
6 mm × 4 mm	1%	30 μL
5.6 mm × 2.4 mm	1%	25 μL
5 mm × 3 mm	1%	30 μL
4.8 mm × 1.6 mm	1%	20 μL
4.5 mm × 2.5 mm	1%	25 μL
4 mm × 2 mm	1%	20 μL
3.5 mm × 1.5 mm	1%	15 μL
3.2 mm × 1 mm	1%	10 μL

lists the adjusted electrode widths and threshold for each pipe size, as well as their respective sensitivities to different volumes of air bubbles in saline.

The overall measurement setup is shown in Figure 9, and the monitoring of power consumption is shown in Figure 10.

5. Conclusions

For the smart infusion device based on IoT technology, bubble detection is a crucial component. This article elaborates on a novel capacitive sensing method that uses only two copper foil electrodes as sensors to determine the presence of larger air bubbles in the infusion pipe by reading the touch value, which varies with capacitance change between the two electrodes through an ESP32 microcontroller. Compared to the commonly used low-cost optical methods, this approach still offers significant cost advantages and some benefits in power consumption. Research results reveal that while the capacitive method exhibits lower accuracy than low-cost optical solutions for regular infusion pipes (6.4 mm × 3.2 mm), it nonetheless satisfies the fundamental design requirements for effective bubble detection. Its remarkably low implementation cost (less than $0.06) and lower power usage (5.18 mW on average) highlight its viability as a cost-effective alternative in the development of smart infusion systems. Applying this method to low-cost IV infusion monitoring devices based on the ESP32 IoT platform can further reduce costs, making product pricing more affordable. Moreover, the capacitive bubble detection method for different dimensions of infusion pipes also ensures a certain level of portability. This method can be applied to most sizes of infusion pipes without fundamentally changing the design and with minimal cost changes. In contrast, traditional optical methods require redesigning the optical path for different dimensions of infusion pipes, and the costs may increase. In some other IoT medical device applications of capacitive sensors, AD7150 is used to read capacitance values at a higher resolution [24,25]. It can directly read the exact capacitance values and achieve a sampling rate of 100 Hz and a resolution of 0.8 fF, which is far superior to ESP32 touch IO ports for reading. It also allows us to use shielded cables for connection, further reducing interference. Additionally, signal processing techniques are applied to analyze the waveform of the capacitance values for more precise determinations. Using a vector network analyzer (VNA) to measure the network parameters of sensors in different scenarios, rather than simply using an LCR meter, can effectively establish a sensor model and might improve sensor accuracy. In oscilloscope measurements, this method effectively reduced environmental noise from 25 mVp-p to 3.033 mVp-p. However, after adding this shielding, the uncertainty in high-frequency measurements on the LCR meter increased. We believe that using VNA to measure transmission and reflection parameters might help determine the reason. These methods might increase costs to some extent but can achieve higher accuracy. For colloidal fluids, the threshold may need to be calibrated individually. Also, this method may not be effective for all types of colloidal fluids.