Software plays a vital role in advancing synthetic biology by enabling the design of biological systems, automating lab processes, managing collaborations, and analyzing large datasets. It is essential for streamlining and improving the engineering of biological systems. iGEM has also been looking into the potential of software since 2006 with the Software and AI village being introduced in 2022.
In our project, we have developed a biosensor designed to detect the concentration of specific biomarkers in people's blood which correlates with the risk of getting cardiovascular diseases. The biosensor employs a colorimetric signal output to indicate different intensities of yellow depending on the concentration of the target. However, to analyze the result, users would have to manually compare the result’s color to a reference card, record their result, and activate the smart braces all on their own. For elders, who are our main target audience, this could be rather challenging. Which is why, to avoid all these hassles and potential errors throughout the process, Digiworm developed a user-friendly phone application (app) that handles everything seamlessly.
Our app plays a significant role in determining whether lumbrokinase is needed to mitigate cardiovascular risk. If lumbrokinase is needed, then the app proceeds to activate the smart braces. Thus, the app is a necessary part in our project to identify the need of using lumbrokinase, connecting our detection kit, or biosensors, to the smart braces.
It currently provides basic functions including a user authentication system, a results checker system, and the most significant of all, our image classification system. We chose to exclude any unnecessary features from our app, as we want our users, primarily elders, to have a simple and effortless experience using it.
We wanted our app to be personalized and follow the diverse needs of our users; therefore, we developed an authentication system (Fig. 1). This system is crucial as it enables us to identify each user, helping both us and them keep track of personal data and past results while ensuring privacy. Additionally, it creates opportunities for future enhancements in development.
Fig. 1. The authentication interface of our app.
With the rise of public interest in AI, its applications have become widespread across various fields around the world. In our project, we decided to apply its image classification feature to our app. Image classification helps interpret and extract valuable information from digital images; specifically in our case, we trained our AI to analyze pictures of testing results (Fig. 2), focusing on the color intensity that indicates the concentration of the targets we are detecting. We chose AI for image classification due to its speed and efficiency, as it can process large amounts of digital images, significantly enhancing our analysis. Furthermore, utilizing AI not only saves time but also prevents potential struggles for users.
Fig. 2. Testing results of our detection kit.
We built our app primarily using Thunkable, which is a mobile app development platform that uses the language React Native (Thunkable, 2021). We chose to use this platform to design our app due to its versatility and convenience. More specifically, Thunkable allows us to integrate and collaborate with other platforms, enabling us to create a more comprehensive software by combining various different addons to our main algorithm seamlessly. Inside Thunkable provided functions that enabled us to connect to Firebase (Thunkable Docs, 2024b), facilitating the creation of an authentication system. It also included a web API and viewer function (Thunkable Docs, 2024a), allowing us to implement features for users to monitor their results. Most importantly, it permitted us to integrate the image classification system we developed into a function of our app. Nevertheless, Thunkable offers great convenience during development, allowing us to design and code our app while instantly viewing the results on mobile devices.
The algorithm of the most crucial function in our app, the image classification system, is written in Python and utilizes the Tensorflow library, along with Keras for building the neural network. We initially developed our image classification system using Teachable Machine, a web-based graphical user interface tool for creating custom machine learning classification models (Chen et al., 2020). By feeding the AI groups of pictures that display seven different color intensities of the testing results, we have trained it to accurately differentiate between various levels of the results (Fig. 3). After this, we exported both our training model and our testing model into Python code for further development (Fig. 4), where we modified the testing model into a cloud database that can be linked to our app coded in Thunkable through its web API function.
Fig. 3. The AI training model we developed for our image classification system using
Teachable Machine.
Fig. 4. Our AI model in Python
Our first version of the app was completed after successfully integrating our image classification system as a feature within it; therefore, we proceeded to test its overall performance. During the testing process, users were able to log into the app, navigate to the camera function, and upload their testing results without any issues. Most importantly, the app was able to analyze and classify the results, informing the user of their current status and whether their result exceeded the threshold (Fig. 5). This confirms that our app is functioning as we intended.
Fig. 5. The results users see after our app finishes the analyzing process.
To determine if you need to take lumbrokinase using the phone application, begin by downloading it from your device's app store. After installation, create an account, or sign in if you already have one, by providing an email and setting your own password. Now, navigate to the camera function of the application to capture an image of your testing result. The app will display your results shortly, including cardiac risk level and whether you need lumbrokinase or not. If the results indicate that lumbrokinase is needed, you may wear the smart braces and wait for lumbrokinase to be released. Additionally, the application also helps track your past results and data, enabling you to monitor your progress over time. Therefore, if you are interested in reviewing any of your past records, you should be able to easily find them on the homepage once logged in.
There is significant potential for our app, and Digiworm aims to enhance existing functions while adding new features. We plan to improve our authentication and image classification systems, as well as introduce an emergency feature to better assist our users. For more information, check out on our Future Plan!
