Explore our project where we developed firmware with ML capabilities for a smart autoinjector and integrated it with our custom app, revolutionizing self-medication and comprehensive care. In collaboration with a leading pharmaceutical company, our team has developed a system that transcends traditional paper logs and manual reminders, fostering a more intuitive treatment experience. This initiative not only focuses on drug manufacturing but also on delivering complete care solutions, including smart packaging that tracks medication intake and supports remote monitoring for a diverse healthcare audience. Our role was integrating various smart devices into their digital ecosystem.
We developed this framework to allow easy creation of customized mobile and web applications on a low-code platform. This modular approach strengthens the connection between patients, their medications, and healthcare providers. We invite you to explore the integration of the smart autoinjector into this system, which is part of a wider suite of smart packaging and application options, including smart inhalers, spirometers, dispensers, among others, all tailored to meet the specific needs of various healthcare providers and their patients.
Challenges
We encountered two distinct yet interconnected challenges related to the smart autoinjector:
Firmware Development for Smart Autoinjector: We had to develop a firmware from scratch with one unique requirement: detection of a moment when a patient injects a drug using a mechanical syringe without any sensors. The microcontroller was only on the holder of this syringe, and together they formed the autoinjector. We could only use the acoustic signals emitted by the syringe at the start and end of the injection. To achieve this, we implemented a neural network.
SDK Development for Device Integration: We were tasked to develop an SDK that enables third-party developers to integrate the smart autoinjector into various healthcare systems seamlessly. It also had to provide a robust and adaptable interface for Bluetooth communication and data management, catering to diverse applications on the main mobile platforms.
Next, we’ll go over the technical solutions and smart strategies we used to overcome these challenges.
Solution for Connected Autoinjector Device
Our solution brings together a smart auto-injector system and mobile apps on iOS and Android. The autoinjector is no ordinary syringe. It has two primary components: a mechanical syringe of medicine and a holder equipped with a microcontroller and skin contact sensor.
When a patient administers an injection, the syringe produces a distinct acoustic “click” to signify the start and end. The app captures data from the injector via Bluetooth, recording the time and date of each injection. If there’s a hiccup in the Bluetooth connection, the smart device ensures no data is lost, storing it to sync later. The app also serves as a reminder system, prompting patients about their upcoming medication doses. As soon as the application receives the injection data, it securely sends this information to the backend server. Doctors can then use a specialized web platform to monitor patient medication intake, view adherence rates, potential overdoses, and more. Immediate alerts, such as deviations in medication intake, are relayed, ensuring timely interventions and consistent treatment outcomes.
Developing Machine Learning for Medical Device
The autoinjector, with its dual-component design, pairs an advanced mechanical syringe with a multifunctional holder. Recognizing the importance of patient feedback, the engineers who built this model of syringe incorporated a mechanical mechanism that produces an acoustic click to signal both the start and end of the injection process. When our team was tasked with developing the firmware for the holder’s controller, we faced a challenge: How can we ensure the holder is aware of the injection process without altering the existing syringe design? As a solution, we equipped the holder with a microphone that was integrated into the firmware we developed, specifically designed to detect the syringe’s clicks. But amidst the ambient noise, recognizing these distinct clicks could be a challenge. We have set up a neural network in the hardware of the holder that is trained to detect the unique sound of these clicks. The training of this neural network involved a meticulous process of data acquisition and processing. A comprehensive sound database was created using a built-in microphone to record the acoustic click of the syringe under various conditions. This encompassed different environmental noises and subtle variations in the click’s sound due to potential damping effects when the auto-injector contacted the skin at various angles. To facilitate efficient data transfer, the Bluetooth firmware was enhanced for increased performance, allowing seamless transmission of the audio stream. Once the data were collected, they underwent a labeling process to distinguish the first and second clicks produced by the device at the start and end of the injection process. This labeled dataset was then fed into TensorFlow, the training algorithm used to develop a sophisticated model capable of accurately recognizing these clicks. The resultant model was subsequently uploaded back into the device. The final step involved rigorously testing the model’s performance and accuracy in click recognition. As soon as the neural network recognizes the sound of a click in the audio signal stream coming from the microphone, it is recorded as the fact of taking the medication.
During the injection, data is also collected and recorded from the skin contact recognition sensor. This sensor has the ability to differentiate between touching the skin and contacting other types of surfaces. Additionally, this sensor enables us to ascertain the angle at which the autoinjector is positioned relative to the skin’s surface during the injection process. As a result, the application receives information not only about the actual execution of the injection but also about the accuracy of its performance.
In the development of firmware for the Nordic Semiconductor nRF52840 microcontroller, used in our syringe holder, our engineers employed Segger Embedded Studio. This integrated development environment (IDE) provides a comprehensive suite of tools for development, including robust debugging and code analysis features. These tools facilitate the efficient identification and correction of errors, and aid in enhancing the performance of the software. With these resources, our engineers are equipped to precisely implement both low-level and high-level firmware functions.
Developing SDK for Connected Medical Device
To meet the integration requirements of our customer’s partners, who aim to incorporate smart devices within their own systems, we have developed a Software Development Kit (SDK) that includes comprehensive Bluetooth communication capabilities. This SDK is a library that streamlines the development process by abstracting the complexity of direct Bluetooth device communication. It empowers developers to focus their efforts on other aspects of the application development, rather than on the underlying Bluetooth communication protocols. We have designed the SDK to be compatible with the leading mobile platforms, utilizing Kotlin for Android and Swift for iOS.
Applying Low Power Design Methodologies
Our engineering team has identified several key strategies for low-power device design, crucial for extending the lifespan and efficiency of battery-powered electronics. An issue they’ve tackled is the voltage drop in batteries, which can affect device performance during power-intensive tasks. For a more in-depth discussion on these strategies, including a progressive wake-up routine that eases the power load on the battery, and other nuanced techniques that ensure power-efficient operation, we encourage you to visit our detailed blog post. This resource is invaluable for teams looking to optimize low-power devices for longevity and reliability.
Moving Forward Together
Seeking a healthcare technology partner? Our expertise in secure, compliant healthcare applications, artificial intelligence, and IoT integration can assist you. We specialize in custom mobile and web solutions that adhere to HIPAA standards, with a focus on streamlining healthcare processes and enhancing data security. Let’s collaborate to innovate! Reach out to begin!
