Team:TUDelft/Hardware

Team

Team Members Collaborations Attributions Sponsors

Project

Project Description Design Engineering Success Results Hardware Contribution Sequencing analysis Partnership Sustainable Development

Lab

Protocols Notebook Parts

Model

Model

Human Practices

Human Practices Integrated Human Practices Education & Public Engagement Proposed Implementation Safety Entrepreneurship

Achievements

Judging Form Medal Criteria Awards and Nominations

Hardware

To ensure inclusivity for our test, an inexpensive and user-friendly device measuring absorbance is required. Here, we present an accessible and robust dedicated read-out device. We were able to build a prototype and assess its usability and functionality.

Introduction

With AptaVita, we aim for a vitamin test that is accessible to as many people as possible, contributing to increasing data availability on vitamin deficiencies and tackling hidden hunger. To make our test available at the point of care, we developed a frugal read-out device that measures absorbance over time to quantify the color-based outcome of the AptaVita test. The read-out device is displayed in Fig. 1. This user-friendly device enables medical personnel with only basic training to take the test. By integrating the input of experts and a user-feedback survey, the design was improved to fit the needs of the user. As the device is portable and standalone, it can be easily used in a variety of settings. The results are stored internally and displayed immediately after the test is finished. With a precise control of light, we ensure that the measurements are robust, as shown in our proof of concept. The stored data can be anonymously uploaded to external databases using a Wi-Fi connection. To increase the accessibility to our test, we took care to keep the costs of the device low (see Entrepreneurship page). The design of the hardware is open-source so anyone can build and operate the device. The complete building guide is available on the Contribution page.

Fig. 1 Prototype of the read-out device.

The electronics are housed in a 3D printed embodiment. A test cassette is slid into a small, closed-off compartment within the embodiment: the test box (Fig. 2). The single-use test cassette carries four paper discs on which the samples can be tested. The user interacts with the device via a touch-screen on top of the embodiment. The progress and results are displayed on this screen.

On this page, the design process of our read-out device is described. To create an impactful read-out device, we adjusted our design choices to the values of our identified stakeholders. Guided by these design choices, we realised a first prototype. In our proof of concept, we showed that our prototype is able to measure different chlorophenol red (CPR) concentrations. Furthermore, we performed user tests to assess the user-friendliness and usability of our device.

Fig. 2 Visualisation of the insertion of the test cassette into the device. The single-use test cassette carries four paper discs that contain the AptaVita system. After blood plasma is added, the test cassette is inserted in the opening of the device, as indicated by the purple arrow.

Integrated Design

Human practices played an essential role in creating an impactful device. The insights and knowledge obtained from stakeholders and experts shaped our read-out device and guided our design decisions. We will go over major considerations, here, for further considerations and more background, one can visit the Integrated Human Practices page.

Design considerations

Based on interviews and evaluations of stakeholder values, the following aspects (privacy, connectivity, robustness, accessibility, and portability) were important considerations that we integrated into the design:

• Privacy: The misuse of health data could give rise to discrimination thereby determining someone’s eligibility for employment, housing or other services. Data storage on personal phones and computers increase this risk because these platforms are well-suited for replication and spreading of data [1]. For more information on this, visit Safe by Design on the Safety page. Initially, we wanted to use personal phones to measure the color change and store the data with an app. After an interview with Dr. ir. Jan-Carel Diehl we changed our design to a dedicated device to mitigate this risk [TU Delft, interview Dr. ir. Jan-Carel Diehl].
• Connectivity: One of the main goals of our project is to increase data availability. This can be made possible by storing measured data in a database, such as the Vitamin and Mineral Nutrition Information System (VMNIS) [1]. This is a database of the World Health Organization (WHO) that contains information on micronutrient deficiencies across the world. In an interview with Dr. ir. Jan-Carel Diehl, we learned that the data should be digital from the start and easy to share. Otherwise, the data is not likely to be uploaded to a database at a later moment. Hence, we decided on a digital readout and set the requirement that our device should be able to share its acquired data anonymously with external databases.
• Robustness: An accurate and reliable diagnosis is considered a minimum requirement for rapid diagnostic tests for health organizations like the WHO and the Centers for Disease Control and Prevention (CDC) [2]. We considered a read-out by eye or with the help of a personal phone camera. However, both options would increase the variability of the read-out [TU Delft, interview Dr. ir. Jan-Carel Diehl]. To ensure low variability, high accuracy and good reliability we decided to go for a dedicated read-out device, as this allows us to control the lightning and temperature.
• Accessibility: We aim for the device to be used in low-resource settings and be accessible for everyone. To make our device accessible, the affordability of the device should be taken into account. Additionally, universality is needed to guarantee the accessibility to our test. Hence, we designed our user interface and the device in a way that anyone can intuitively understand how to work with it. More information about the design of our user interface can be found here . The user feedback that we obtained after building our device can be found here.
• Portability: To use the device at the point of care, the device needs to be portable. A healthcare worker should be able to carry the device without additional equipment or people and conduct the test without the need of external power.

Morphological chart

The morphological chart shown below provides a deeper insight into the design choices made to optimize our product. A morphological chart is an overview to capture the necessary product functionality and explore alternative means and combinations of achieving that functionality [3]. In the first column, we describe all required functionalities, and in the second, third and fourth columns, we propose different ways of achieving this functionality. The concept that best fit the functional requirement is highlighted. In the fifth column, an explanation is given for why a specific concept is chosen.

Morphological chart

Tab. 1 Morphological chart

Functional requirements	Concept 1	Concept 2	Concept 3	Remarks
Accurate measurements	An LED with a specific wavelength and a sensor	A camera with a colour filter	An LED and a sensor with a colour filter	We use a LED with a specific wavelength and a sensor, due to its affordability and simplicity. A camera would have been too expensive and a colour filter would increase the complexity of the device.
Consistent lighting	Dark box with a light source	Using sunlight and calibrate before testing	Conduct the test in a dark room	We use a dark box with a light source to ensure a robust read-out. We do not use sunlight or a dark room as the light intensity varies more and the flexibility of use at the point of care decreases.
Temperature control	A piezoelectric element that can cool and heat	Resistors that can heat	A temperature-controlled room	We use resistors as a compromise between costs and energy consumption and the ability to cool of the piezoelectric element. Heating with resistors is also more precise and flexible than a temperature-controlled room.
Safe and easy handling of the paper discs	A one-time use test cassette that can be slid into the device	A one-time use test cassette that can be put in a part of the device that is closed off with a removable lid	A one-time use test cassette that can be put on a tray like a CD	We chose for sliding the test cassette into the device, as this requires fewer moving parts of the device and less actions by the user compared to the other options.
Easy to operate	An LCD touchscreen	LCD screen controlled by buttons	LEDs (red/green) controlled with buttons	We use an LCD touchscreen to improve the universality of the device, as the touchscreen provides more information than LEDs and is more user-friendly than a regular screen.
Stability	Low center of mass	Rubber at the bottom of the device	Suction cups at the bottom of the device	We chose for only lowering the center of mass to minimize the complexity and costs. If this is insufficient, the other concepts can be implemented in future prototypes.
Easy to produce	3D printable components	Create the components from a plastic mold	A foldable device with a cardboard embodiment	We 3D printed the components as this is easy, robust and affordable, and allows easier reproduction of the device by other iGEM teams. A plastic mold would be more efficient for mass production.
Connectivity to database	Wi-Fi connection to upload the data to a database	Bluetooth or Wi-Fi connection from the device to a phone/ laptop, from which the data is transferred to a database	Use a physical connection, e.g. an USB cable, to upload the data	We chose for a Wi-Fi connection to upload the data, as this is simpler than a physical connection or a connection to another device.
Material	Metal	Plastic	Cardboard	We use plastic as it is affordable, durable and safe to use for medical devices [4], and it keeps the device light enough to be portable.
Power supply	Use a power bank	Connect the device to a solar panel	Use a wall plug	We use a power bank to power our device so that it is portable, and can be used easier at the point of care.
Information processing	Raspberry Pi	Arduino	Attach to laptop	A Rapsberry Pi has sufficient computing power and memory, as opposed to the Arduino. The Raspberry Pi does not require external devices and is thus simpler than the laptop.

Prototyping

In the integrated design, concepts were selected that best met the functional requirements of the device. Guided by these concepts, we realised a first prototype. The device consists of four main plastic parts: a test cassette, a test box in which the test takes place, an outer case, and a power bank holder. The test cassette supports four paper discs that contain the AptaVita test. The casette keeps the paper discs in the correct position, and can be slid into the test box. This box is located inside the outer case. The light and temperature are controlled in the test box, allowing for consistent measurements. The outer case protects the components and electronics. The power bankholder is placed at the bottom of the outer case. The power bank supplies the power to the Raspberry Pi 4B and the electronic components. The inside of the device is shown in Fig. 3.

Fig. 3 A look inside the read-out device. The test cassette on the right is slid into the test box at the top part of the outer case. The Raspberry Pi and boards on which the electronics are soldered can be seen in the bottom part of the outer case.

During a test, a paper disc is repeatedly illuminated with yellow light and the absorbance of the paper disc is measured. The color of the paper disc will change over time as chlorophenol red-b-D-galactopyranoside (CPRG) is converted into CPR. CPR is purple and absorbs mainly yellow light. Hence, an increase in CPR over time will lead to an increase in absorbance over time. The relation between the absorbance and the concentration is described by the Beer-Lambert law:

where A is the absorbance, ε the extinction coefficient, l the optical path length, and c the concentration. In general, this law is valid for low to moderate concentrations of the absorbing species [5]. The rate of CPR production can therefore be derived from light intensity measurements, which can be linked to the vitamin concentration. Fig. 4 provides a schematic overview of the measurements over time. A more detailed explanation of our system can be found on the Design page.

Fig. 4 Schematic representation of the change in absorbance over time of a paper disc. The paper disc is illuminated and absorbs a fraction of the light. As the CPR concentration increases over time, the disc becomes more purple and the absorbance increases. The intensity of the transmitted light is measured with an ambient light sensor.

Measuring absorbance

To measure the absorbance of the paper discs, we designed a single-use test cassette that carries the discs (Fig. 5). The user has to add the blood plasma on the paper discs and slide the cassette into the test box. The holes above and underneath the discs allow light to pass through the disc, so absorbance can be measured. As a single test cassette carries four paper discs, multiple vitamins can be tested at the same time. For this prototype, we made a reusable cassette to minimize waste and costs. However, for the final implementation of the hardware, we envision single-use test cassettes to avoid contamination.

Fig. 5 Image of the reusable test cassette.The paper discs are clamped between the bottom and top part of the cassette. The light passes through the round holes, and air can flow through the rectangular holes. (a) The two parts of the test cassette before assembly. (b) An assembled test cassette.

In the test box, each paper disc is illuminated from above by an LED. These LEDs emit a wavelength range of 550-600 nm with a peak at 574 nm [6]. This peak corresponds to the absorbance peak of CPR [7]. The four LEDs are sequentially switched on and off (Fig. 6). Therefore, only the LED directly above the paper disc is on when the absorbance is measured of that disc. The electronic circuits are designed to ensure that the intensity of the LED is stable. The electric circuit is described in the additional information below.

Fig. 6 Video of the LEDs switching on and off one at the time.

A light sensor is placed below the hole of each paper disc. This light sensor allows for a quantitative measurement of the intensity of the transmitted light. To protect the light sensor against liquid and dust, a layer of transparent polycarbonate is placed between the light sensor and the test cassette. During a test, the absorbance of each paper disc is measured sequentially. This sequential measurement is continually repeated until the test is finished. The first measurement for every paper disc is taken as a blank. The absorbance A is calculated by

where I₀ is the intensity of the blank (first measurement), and I is the intensity of a measurement that is not the blank.

Additional information about the absorbance measurement

To ensure that the LEDs illuminate the discs with a constant intensity, a voltage-to-current circuit was used (Supplementary Fig. 1) [7]. This circuit regulates the current through the LED, and the intensity of the emitted light is proportional to the current [8]. The 3.3 V output of the Raspberry Pi is connected to a resistive voltage divider consisting of R_16-19, of which the output voltage is connected to the positive input of the opamp. The opamp regulates its output voltage such that the voltage at the positive and negative input is equal. This will be the case if the voltage drop over resistors R_20-21 is equal to the positive input voltage. This depends on the current by Ohm’s law: V_21-22 = I_21-22 R_21-22. Therefore, the base current that the opamp will supply to the transistor will have a value such that the collector current equals I_21-22. As the magnitude of I_21-22 only depends on passive components, namely resistors, and negative feedback is supplied to the opamp, I_21-22 will be stable [9]. By putting the LEDs in series with R_21-22, the current I_21-22 (minus the base currents of the transistors) will consistently flow through the LEDs.

Supplementary Fig. 1 Electrical diagram for the control of the LEDs.. The LEDS are switched on individually with transistors, which are connected to the Raspberry Pi. The current through the LED is regulated by a voltage-to-current circuit. For a full overview of all electrical diagrams, see Downloads. In this diagram, the GPIO pins of the Raspberry Pi 4B are also specified.

Communication between the sensors and the Raspberry Pi takes place with the I²C-protocol [10]. For this protocol, it is required that the different sensors have unique addresses. To achieve this, the sensors were connected to the Raspberry Pi via a multiplexer. The multiplexer enables the selection of a single sensor.

Blocking ambient light

For the measurements, we needed to ensure that the test box is opaque to minimize the ambient light coming through the test box and reaching the sensors. For conventional 3D printers, the infill percentage can be set by the user [11]. This infill percentage is visualized in Fig. 7. Decreasing the infill percentage lowers the print time and conserves material, while still providing strength. However, as the density of plastic is lowered, the walls become more translucent.

Fig. 7 Picture that shows different levels of infill percentage for 3D-printing. The infill percentages shown are 12%, 30% and 50% from left to right. The higher the infill percentage, the higher the density of the plastic is, as depicted in this figure adapted from Filament2Print [12].

The test box was sprayed with black matte spray to create an opaque test box. The inside of the outer embodiment was also spray painted to minimize incoming reflection of light entering the device. The outside of the embodiment was not spray painted so it is clearly visible if the device needs to be cleaned. To test if the ambient light is sufficiently blocked, the ambient light intensity was measured with the device in four conditions. These conditions are summarized in Tab. 2, and photographs of these conditions are depicted in Fig. 8. The experiment was performed in a brightly lit room. The measurements from each sensor were averaged over four minutes to account for fluctuations in the incoming light. Fig. 9 shows that already approximately 80% of the light was blocked by the test cassette (condition Fig. 8b), compared to a completely opened device (condition Fig. 8a). Closing the test box reduces the light by 98% for every sensor. When the embodiment was also closed, the sensors did not detect any light. Hereby, we show that the ambient light is blocked effectively.

Tab. 2 Overview of the test conditions of the ambient light blocking experiment.

	Closed embodiment	Closed test box	Test cassette in test box
Condition 1	-	-	-
Condition 2	-	-	+
Condition 3	-	+	+
Condition 4	+	+	+

Fig. 8 Photographs of the four conditions of the ambient light blocking experiment. The sensors are positioned below the holes in the black test box at the right side of the device. The inside of the embodiment, the test box, and the test cassette were spray painted with black matte spray to prevent light pollution from outside. (a) The embodiment and test box were both opened. The test cassette was not placed on the bottom half of the test box. (b) The embodiment and test box were both opened. The test cassette was placed on the bottom half of the test box. (c) The embodiment was opened, but the test box was closed. The test cassette was slid into the test box. (d) The embodiment and the test box were both closed. The test cassette was slid into the test box.

Fig. 9 Percentage of transmitted light relative to the light intensity that was measured with an opened device. The light intensity was measured in four different conditions, which are summarized in Tab. 1. For each sensor, the percentage of transmitted light was calculated by normalizing with respect to condition 1. The measurements of a sensor were averaged over four minutes to account for fluctuations in the incoming light. The standard error of the mean was lower than 0.12% for all conditions and sensors.

Regulating temperature

We aimed to regulate the temperature of the test box at 37 °C since the reaction rates of the cell-free system are temperature-dependent. This would make the test more robust, because the variability of the test conditions would decrease. Before a test cassette is inserted and the test is started, the box should already be pre-heated to 37 °C to ensure that the results are as consistent as possible. High wattage resistors are placed inside the test box and function as heating elements to heat up the test box. To allow for air circulation inside of the test box and a more homogeneous temperature, we added a ventilation grille in the test cassette (Fig. 5). Additionally, the volume of the test box was minimized and the walls of the test box were 10 mm thick to minimize heat loss through thermal conduction.

At the top of the test box, the temperature is measured by a sensor. Based on this, the heating by the resistors is adjusted over a range of 0 to 3.7 W . We chose the upper limit of the power range based on an estimation of the heat loss. The required root-mean-square power P_RMS is calculated based on the temperature measurement in the following way:

where K_p is the proportional gain, K_i is the integral gain, K_d is the derivative gain, and e(t) is the error between the desired temperature and the measured temperature, calculated as e(t) = 37 °C - T(t). The required power is calculated once per absorbance measurement, so roughly once per second. The power is increased both if the e(t) grows (proportional term), and if the difference is present over a longer period of time (integral term). The derivative term is proportional to the rate of change of e(t), so that the power is increased if e(t) quickly increases, and decreases if e(t) quickly decreases. This smoothens the error. This control system, or PID-controller, is one of the most extensively used methods for temperature control, because of its simple implementation and often robust performance [13]. There are different methods to tune the parameters, one of them being the Ziegler-Nichols tuning. This is a heuristic tuning method, in which the K_i set to zero, and K_p is increased until stable oscillations are observed. From the oscillation time and the associated proportional gain, values for K_p, K_i and K_d are calculated [14]. As the PID-controller is a simple and widely applicable controller, we think it may be possible to obtain a steady temperature in the test box using this controller.

During the prototyping of the temperature regulation, one of the components, a transistor, burnt out. This transistor is used to switch the heating on and off, and is regulated by the Raspberry Pi. We think that it burnt out because the base current to the transistor was too low, which would lead to a high voltage drop over the transistor. This would result in heat dissipation in the transistor instead of the resistors. To prevent this, a transistor with a lower base current requirement for switching could be used. Alternatively, the base current could be amplified. Due to time constraints, we were unable to test this.

Additional information about the temperature regulation

To get an idea of the power required to heat the test box, we made an approximation of the heat loss of the test box. We assumed that the temperature of the test box is homogeneous and the walls are solid, and we neglected the holes in the test box. The heat flow H through the walls of the box is then given by

where k is the thermal conductivity, A the surface area, ΔT the temperature difference between the walls, and Δx the width of the walls. The walls have a width of 1 cm, and are made of Polylactic Acid (PLA), which has a thermal conductivity of k = 0.185 W m^-1 K^-1 [15]. For the surface area, the area of the inside of the box was used, which gives A = 7966 mm². For the temperature difference, we assume that the embodiment and room temperature are both 20 °C, so that ΔT = 17 °C. We expect that the air of the embodiment is warmer, but we neglect this for this approximation. Using these numbers, H = - 2.5 W.

To heat the box, power resistors are placed in the test box. These resistors are connected to the 9 V output of the power bank in parallel. The effective resistance of the circuit is equal to R_eff = R n^-1 = 220 / 10 = 22 Ω, where n is the number of resistors and R the resistance of one resistor. The power that can be dissipated is given by P = V² R_eff^-1 = 81 / 22 = 3.7 W. This is above the approximation of the heat loss, including some margin.

The temperature is measured with a sensor at the top of the test box. The resulting analog signal is processed by an analog-to-digital (ADC) converter, which transmits the data to the Raspberry Pi using the Serial Peripheral Interface (SPI) protocol (Supplementary Fig. 2).

Supplementary Fig. 2 Electrical diagram for the temperature read-out. The output of the LM35DZ temperature sensor is amplified with a non-inverting amplifier circuit. The output of that circuit is connected to the MCP3008 ADC-converter, which transmits the data to the Raspberry Pi. For a full overview of the electrical diagrams, see Downloads.

Powering the device

The device is powered by an 18 W power bank. The power bank housing is located on the bottom of the embodiment and is not integrated into the primary embodiment as can be seen in Fig. 10. This location lowered the center of mass of our device and it allows for easy access to the power bank. This design was chosen to allow the user to remove an empty power bank and replace it with a fully charged power bank. However, after an interview with Dr. ir. J.C. Diehl, it became apparent that a balance should be made between repairability and theft avoidance. A power bank can be used for a whole lot of other devices and would be interesting to take home or be sold [TU Delft, interview Dr. ir. Jan-Carel Diehl].

Fig. 10 Power bank housing located at the bottom of the device. It was placed at the bottom to lower the centre of mass.

User interface

We created a distinct user interface to ensure accessibility and universality. Our interface centres around a home screen which has three options: Test, Upload Data and Settings. When a user wants to run a test, the Test option should be chosen. The user will then be guided through all the steps required to perform the test. Once the test has been completed, the results will be shown on the screen. The data is stored on the device, so that the health care worker can upload the data at a suitable moment. This can be done via the upload data option on the home screen. The data will be uploaded to a database over Wi-Fi. In the Data section, an overview is given about how much information is stored. Finally, in Settings, the user can select his preferred settings. We included language, Wi-Fi and other general settings. In the carousel below you can scroll through the user interface.

User interface (1/19)

User interface (2/19)

User interface (3/19)

User interface (4/19)

User interface (5/19)

User interface (6/19)

User interface (7/19)

User interface (8/19)

User interface (9/19)

User interface (10/19)

User interface (11/19)

User interface (12/19)

User interface (13/19)

User interface (14/19)

User interface (15/19)

User interface (16/19)

User interface (17/19)

User interface (18/19)

User interface (19/19)

Previous Next

To create a universal and accessible interface, the following considerations were implemented:

• We added the option to change the language of the user interface, so that a user can change it to his local language, which improves the universality of our device.
• We only record the patient's age, gender, and location to protect the privacy of the patients.
• We integrated loading bars so that a user can monitor the progress whilst waiting upon a suggestion by Dr. ir. Jan-Carel Diehl.
• We added smileys to indicate the results using a smiley after Dr. Ria Reis reminded us of the significance of the universality of the readout signs [Interview Dr. Ria Reis]. For the same reason, we changed text to check marks, which was suggested by Dr. ir. Jan-Carel Diehl.
• We made changes to the user interface after a user test, improving its user-friendliness and clearness of the user interface. This includes additions like adding a battery symbol on the top right, adding an option to return to the previous step when starting the test, and renaming the general settings section.

Implementation

Proof of concept for concentration measurements

Our final goal is to measure the CPR concentration over time during a test. The measured concentration curve can be linked to the vitamin concentration of a sample. Here, we demonstrate the ability of our device to measure colorimetric differences between CPR concentrations. The CPR was diluted with Phosphate Buffered Saline (PBS), which has pH of 7.4 This pH mimics the conditions in which the AptaVita test will take place. The reaction takes place in the blood plasma which has a pH of 7.4 and the reaction buffer has a pH of 7.9 [16, 17]. However, as the volume and buffer capacity of the blood plasma are significantly higher than those of the reaction buffer, we expect that the final pH is around 7.4.

To validate that CPR absorbs light in the range of wavelenghts emitted by the LEDs of our read-out device, we measured the absorbance spectrum of CPR on a paper disc in a plate reader. This spectrum (Fig. 11) shows an absorbance peak at a wavelength of 579 nm. The emission peak wavelength of our LEDs is at 574 nm, indicating that the LEDs are suited to measure the absorbance of CPR at wavelengths near its absorbance peak. Measuring the absorbance at these wavelengths results in the highest possible sensitivity, because the same increase in concentration will lead to a higher increase of absorbance compared to the other wavelengths.

Fig. 11 Absorbance spectrum of CPR on a paper disc measured with a plate reader. The disc was hydrated with 5.5 µL of 0.25 mM CPR. The CPR was diluted with Phosphate-Buffered Saline (PBS). A disc with 5.5 µL PBS was used as a blank measurement.

To measure the absorbance at a range of CPR concentrations, we linearly diluted the CPR to 0.25 mM, 0.5 mM, 0.75mM, 1mM and 1.25 mM (see Fig. 12). For each concentration, the absorbance was measured in duplicate in the plate reader, and triplicate in the read-out device. As a blank measurement, we used a paper disc that was hydrated with the same buffer as used for the dilution. In the plate reader, we observe a linear relation with an R² of 0.97 (Fig. 13a). This shows that the Beer-Lambert law is valid for this range of CPR concentrations. In the read-out device, we also observed a linear relation between the absorbance and the CPR concentration with an R² of 0.95 (Fig. 13b). This demonstrates that the device is capable of distinguishing different CPR concentrations through absorbance measurements.

Fig. 12 CPR diluted with PBS. From left to right, the concentrations are 0 mM, 0.5 mM, 0.75 mM, 1 mM, and 1.25 mM.

Fig. 13 Absorbance against CPR concentration on paper discs. The CPR was diluted with PBS. Each paper disc was hydrated with 5.5 µL of the CPR dilution. A linear curve was fitter for both measurements. (a) Absorbance measurements by a plate reader (y = 0.61 x + 0.16, R² = 0.97) (b) Absorbance measurements by the AptaVita read-out device (y = 0.80 x + 0.137, R² = 0.95).

The slope of the fitted curve for the read-out device is steeper than the slope of the fitted curve for the plate reader. We speculate this may partly be caused by a difference in path length between the devices. The path length is proportional to the slope in Fig. 13a by the Beer-Lambert law as A = (εl) · c, where εl determines the slope [5]. Because a relatively high volume of liquid was added to the disc, some of the liquid is not taken up by the disc. The excess liquid can flow to the sides of the well. This lowers the total path length. As the well of the test cassette is smaller than that of the plate reader, more liquid stays on top of the disc in the read-out device, which results in a longer path length and thus a steeper slope.

User feedback

To assess the user-friendliness and usability of our device, we performed user tests. All user tests were performed before the device was taken into the lab for measurements with CPR, and thus all biosafety standards were followed. We invited six participants who were not involved in the project to perform a user test. These user tests provided insight into the interaction of users with our device. It would be of value to perform additional user tests with healthcare workers in the local context, which is an essential step in the development of the device. The participants were seated next to a moderator who gave a general introduction to the project. The moderator asked the participants to perform specific actions, while the moderator observed how the participants interacted with the device. The participants were asked to think out loud during their action, and afterwards, the participants scored the clearness and ease of performing the actions.

Fig. 14 Photos taken during the user test. The participants interact with the device. They comment on their experience, and score the actions on clarity and ease. The actions include (a) navigating the user interface, (b) lifting the device from the table, and (c) inserting and removing the test cassette from the device.

The participants scored the clearness and ease of performing the actions. A score was given on a scale of one to five. A score of one indicated that the action was either unclear and/or hard to perform, and five noted that it was straightforward and/or easy to perform. The survey consisted of two parts. In the first part, the handling of the device was assessed, and in the second part, the interacting with the user interface was assessed. For a more detailed description of the survey and the results per participant, click here. The average score of the six participants for the different actions is shown in Fig. 15.

Fig. 15 Visual representation of the results of the user tests. The participants were asked to perform actions and score these on clearness and ease. The average of the scores is displayed (n = 6), where a score of one indicates that an action was unclear or hard to perform, and a score of five indicates that the action was straightforward or easy to perform.

From the data in Fig. 15 and the interaction with the participants, it became evident that all the actions related to the power bank are complex, especially locating the power bank holder. The actions related to testing cassettes were more straightforward, but some of the participants initially struggled with the orientation of the cassette. Picking up and carrying the device was not a problem for any of the participants.

The user interface was also tested and scored based on its usability. Starting the test was not an issue for any of the participants. Some of the participants would like additional data in the result section, such as error margins. Given that the accuracy of the test must be sufficiently high before the test can be implemented, we think the end-user should be able to rely on the test result being accurate enough without having to assess the uncertainty himself after every test. Uploading data was unclear for some as it was not explicitly mentioned what data was being uploaded. Finally, none of the participants had a problem with changing the language settings of the device.

Based on the feedback obtained from the user test we made modifications and recommendations for the future development of our device and user-interface:

• Tactile feedback: The position of the test cassette is crucial as the paper discs have to align with the LEDs and sensors to ensure an accurate read-out. Two out of the six participants struggled with inserting the cassette and did not position it correctly. To avoid this, we will add a physical response of the device to the user input, also known as tactile feedback. This can be as simple as a click if the test cassette is positioned correctly. This indication will let the user know that the test cassette is positioned correctly.
• Use cues: All six participants struggled with the position and orientation of the power bank and/or the test cassette. To improve the clarity, we have added use cues in the form of color tags to elucidate the position and orientation (Fig. 16). We went with color tags instead of text to maintain the universality of our device. This can also be done for the power bank.
• Integrating power bank into the device: As the actions concerning the power bank were perceived as difficult, it would be better to integrate the power bank into the device, so that the users do not need to perform this action anymore.
• Ventilation grille: One participant also pointed out that the location of the ventilation grille does not protect the electronics. To improve this, we would change the position of the ventilation grille to the side and add a filter to protect the electronics. Additionally, a smaller ventilation grill would suffice if active cooling is implemented.
• Removing yellow category test results: In the results section, the test result is indicated by a smiley and an action. From the user test it followed that the action related to the yellow smiley, accompanied by the text "Fine" was unclear. Considering the fact that a doctor will presumably mainly require if a patient is deficient or the exact vitamin concentration, we removed the “Fine” category. However, to make a final decision about this category, a user test with the local doctors is required.
• Clarity data uploading: It was clear for all participants what actions were required to upload the data. For four out of six participants, however, it was vague what the data represented. To improve this we have decided to add the word ‘patients’ after the number as that explains that this number denotes the amount of patients of which data is stored on the device.
• Minor changes user interface: We also made some minor changes to the user interface, improving its overall user-friendliness and clearness. This includes additions like adding a battery symbol on the top right, adding an option to return to the previous step when starting the test, renaming the general settings section.

Fig. 16 Use cues for position and orientation. After user tests, red stickers were placed on the test cassette and the test box, and yellow stickers were placed on the power bank and power bank holder. By placing the indicator on the upper right side of the test box, the correct orientation of the test cassette is immediately clear.

Future prospects

The dedicated read-out device requires further development to effectively implement it into the real world. The current proof-of-concept shows that different CPR concentrations can be distinguished, but the device should still be optimized to increase the sensitivity and precision. Furthermore, the input that we received from the user tests and experts should be taken into account in the next prototypes. We thought of the following aspects that could be integrated into the design:

• Improving sensor performance: We think there are multiple ways in which the sensitivity of the ambient light sensors can be increased. This will lead to more precise absorbance measurements, so that smaller differences in CPR concentrations can be distinguished. First of all, the positions of the LEDs and the sensor can be improved. The position of the LEDs was not perfectly straight above the wells, because of human error during the soldering. Additionally, the sensor was about 1 cm below the bottom of the test box to make space for some cables. By improving the positioning of the LEDs and reorganizing the cables, the amount of light that reaches the sensor can be increased, so that a larger part of the range of the sensor is used, thereby increasing the precision and sensitivity of the measurements. Additionally, the integration time of the ADC-converter of the sensor could be increased. This would increase the time of a single measurement. As the signal is integrated over a longer timespan, the sensitivity of the sensor is increased [18]. We used an integration time of 101 ms, but the sensor also has the option to use 175 ms or 699 ms. However, when we tried this, we observed unexpected behaviour, namely a sawtooth pattern in the measurements over time. We are unsure of the cause of this behaviour. It should also be noted that a longer integration time, eg. 699 ms, would decrease the number of datapoints per test, which may be problematic in the analysis of the measurement.
• Measurement over time: Although we showed that our device is capable of performing measurements on the same disc over a time period, we have not tested if the device can measure a change of CPR concentration over time. This would be the next step in testing our read-out device. In the end, the measured temporal evolution of the CPR concentration could be compared to reference curves that correspond to known vitamin concentrations. The vitamin concentration of the test sample could be inferred from the comparison.
• Alternative paper-based systems:The current read-out device is designed to measure the absorbance of four paper discs positioned next to each other for the AptaVita test. By adapting the test cassette or choosing LEDs that emit light at a different wavelength, it is possible to measure the absorbance for alternative paper-based tests. An example is using the device to reliably diagnose malaria on a paper-based system [19]. It would only be required to change the LEDs to measure at a different wavelength. Other possible paper-based tests that could use our read-out device include a test for bacteria or a test for the Zika virus [20,21].
• Multiple tests at once: Currently, the device is designed for loading one test at a time. However, the time for one test is approximately 45 minutes. Therefore, Dr. ir. J.C. Diehl recommended developing a device with multiple slots to enable a larger capacity in busy hospitals and test more patients. This would increase the capacity of the read-out device and thereby also reduce the costs.
• Integrate power into the device: In the initial design, the battery pack is placed outside the embodiment as this allows the user to remove an empty power bank and replace it with a fully charged power bank. However, Dr. ir. J.C. Diehl pointed out that this also means that the battery pack gets lost more often and that the battery pack might be valuable to people for other purposes. Future developers could adjust the design of the device to have the battery pack secured in the device.
• Tactile feedback: The position of the test can vary slightly when placing it in the device. To ensure that the test cassette is positioned identically every time, a physical response of the device upon inserting the test cassette can be added. This is also known as tactile feedback, and could be as simple as a click if the test cassette is positioned correctly.
• Active cooling: We use resistors to regulate the temperature as a compromise between cost and energy. It is essential to consider that resistors can only heat the device and not cool it. Although it is not common in Uganda, Sub-Saharan Africa and Southeast-Asia for temperatures to rise above 37 °C, a cooling mechanism could still be integrated into the device to ensure that the test can always run at 37 °C. Another advantage of this is that the ventilation grille can be made smaller, so that the electronics are better protected by the embodiment.
• Manufacturable PCB: For this proof-of-concept, we used perfboards to solder the electronics as this allows us to develop and design the circuits easily. The next step, however, is to integrate the electronics on a printed circuit board (PCB). This will reduce the size of the device drastically, as well as the costs and the production time.

Downloads

The design was open-sourced so that anyone can build and operate the device. We share the files for the 3D printed components and provide a detailed assembly guide to construct the embodiment. Furthermore, the complete electrical diagram is specified, and we described how to solder everything in the correct position. In addition, all the software necessary to operate the read-out device is provided. For the full building guide visit our Contribution page, and for the software to operate the device, visit our Github repository.

References

1. Tiffin, N.,George, A.& LeFevre, A.E. (2019). How to use relevant data for maximal benefit with minimal risk: digital health data governance to protect vulnerable populations in low-income and middle-income countries. BMJ Global Health [https://gh.bmj.com/content/bmjgh/4/2/e001395.full.pdf]
2. Land, K. J., Boeras, D. I., Chen, X. S., Ramsay, A. R., & Peeling, R. W. (2018, December 13). REASSURED diagnostics to inform disease control strategies, strengthen health systems and improve patient outcomes. Nature Microbiology, vol. 4, 13 Dec. 2018, pp. 46–54.https://www.nature.com/articles/s41564-018-0295-3
3. Börekçi, N. A. (2018). Design Divergence Using the Morphological Chart. Design and Technology Education, 23(3), 62-87.
4. Czuba, L. (2014). Application of Plastics in Medical Devices and Equipment. Handbook of Polymer Applications in Medicine and Medical Devices, 9–19. https://doi.org/10.1016/b978-0-323-22805-3.00002-5
5. Ball, D. (2006). Field guide to spectroscopy. Bellingham, Wash.: SPIE Press.
6. Kingbright. (2019). WP7113CGCK [Datasheet]. Retrieved from https://eu.mouser.com/ProductDetail/Kingbright/WP7113CGCK?qs=%2Fha2pyFadugjtNsyFL6YAh1DgmyDFXNsj6NJZ3M3NUE=
7. McNerney, M. P., Zhang, Y., Steppe, P., Silverman, A. D., Jewett, M. C., & Styczynski, M. P. (2019). Point-of-care biomarker quantification enabled by sample-specific calibration. Science advances, 5(9), eaax4473.
8. Bastemeijer, J. (2019). Theory electronics. Lecture, Delft.
9. Storey, N. (2017). Electronics (pp. 265-275). [S.l.]: Pearson Education Limited.
10. Hofei90. (2020). GitHub - Hofei90/python-apds9930: Python library for the APDS-9930 I2C Ambient Light and Proximity sensor. Retrieved 10 September 2021, from https://github.com/Hofei90/python-apds9930
11. O'Connel, J. (2021). 3D Printing Infill: The Basics – Simply Explained. Retrieved 11 October 2021, from https://all3dp.com/2/infill-3d-printing-what-it-means-and-how-to-use-it/
12. Filament2Print. (2021). Image 1: Different infill percentages [Image]. Retrieved from https://filament2print.com/gb/blog/71_importance-infill-3d-printing.htm
13. Tan, N., & Atherton, D. (2006). Design of stabilizing PI and PID controllers. International Journal Of Systems Science, 37(8), 543-554. doi: 10.1080/00207720600783785
14. PID Tuning via Classical Methods. (2021, March 5). University of Michigan. https://eng.libretexts.org/@go/page/22413
15. Laureto, J., Tomasi, J., King, J., & Pearce, J. (2017). Thermal properties of 3-D printed polylactic acid-metal composites. Progress In Additive Manufacturing, 2(1-2), 57-71. https://doi.org/10.1007/s40964-017-0019-x
16. Aoi, W., & Marunaka, Y. (2014). Importance of pH homeostasis in metabolic health and diseases: crucial role of membrane proton transport. BioMed research international, 2014, 598986. https://doi.org/10.1155/2014/598986
17. New England Biolabs. T7 RNA Polymerase. Retrieved October 3, 2021 from https://www.bioke.com/webshop/neb/m0251.html
18. Avago Technologies. (2015). APDS-9930 [Datasheet]. Retrieved from https://nl.farnell.com/broadcom/apds-9930/light-proximity-sensor-digital/dp/3677127.
19. Lathwal, S., & Sikes, H. D. (2016). Assessment of colorimetric amplification methods in a paper-based immunoassay for diagnosis of malaria. Lab on a Chip, 16(8), 1374–1382. https://doi.org/10.1039/c6lc00058d.
20. Safavieh, M., Ahmed, M. U., Sokullu, E., Ng, A., Braescu, L., & Zourob, M. (2014). A simple cassette as point-of-care diagnostic device for naked-eye colorimetric bacteria detection. The Analyst, 139(2), 482–487 https://doi.org/10.1039/c3an01859h.
21. Pardee, K., Green, A. A., Takahashi, M. K., Braff, D., Lambert, G., Lee, J. W., Ferrante, T., Ma, D., Donghia, N., Fan, M., Daringer, N. M., Bosch, I., Dudley, D. M., O’Connor, D. H., Gehrke, L., & Collins, J. J. (2016). Rapid, Low-Cost Detection of Zika Virus Using Programmable Biomolecular Components. Cell, 165(5), 1255–1266. https://doi.org/10.1016/j.cell.2016.04.059.

A big thank you to our sponsors!