InfuLink

What it does

InfuLink uses a Raspberry Pi and AI camera to capture syringe pump images. Generative AI extracts data in real time, which is shown on a web-based dashboard and synced with Electronic Medical Record (EMRs) to reduce workload and improve healthcare efficiency.

Your inspiration

Inspired by the burden on nurses observed in the ICU at UMMC, InfuLink was developed to reduce manual syringe pump monitoring and data recording. Many patients require multiple pumps, leaving nurses overwhelmed. Despite advances in healthcare, most pumps still lack EMR connectivity and automation. Smart pumps are costly and often incompatible with hospital systems. InfuLink offers a practical, affordable solution by capturing pump data in real time, storing it for traceability, and syncing with EMRs. As our slogan says—We link. We care.

How it works

InfuLink is a smart external add-on that transforms traditional syringe infusion pumps into connected devices with real-time monitoring and EMR integration. It features a 3D-printed adjustable frame that holds electronic components, including a Raspberry Pi and AI camera, which captures high-resolution images of the pump display. A barcode scanner links each pump to the correct patient. Images are processed using the Gemini 2.0 Flash-Lite AI model to extract infusion data like flow rate, dosage, and volume infused. Data is stored locally and shown on a live web dashboard with alerts and historical logs. The dashboard offers three views: a hospital-wide overview, a ward-specific page, and a detailed patient view. It syncs with OpenEMR for automatic updates, reducing manual work, increasing accuracy, and improving care efficiency while adapting to current hospital systems.

Design process

We began our design process by identifying key healthcare challenges through interviews with medical practitioners, nurses and engineers. These discussions provided insights on the current needs of syringe pumps, especially on automated data logging and remote monitoring capabilities. With that, we brainstormed multiple concepts including direct serial communication and camera-based detection. After evaluations, we opted for a camera-based approach using Raspberry Pi and AI camera due to its non-invasive nature and adaptability across syringe pump models. The first prototype included a basic frame with a simple clamping mechanism which is clunky and unstable. Through several redesigns, we made improvements such as adjustable arms, inserting boxes for simple mounting and dismounting, camera sliding mechanism, etc. Software-wise, we originally used OCR models for text detection which we found to be difficult as there are also symbols and figures on the screen of the syringe pumps. Hence, we employed Generative AI models which not only improves data accuracy, but also categorize them to be displayed on our customized web-based dashboard. With each iteration of the design, we focus on improving the mechanical stability, software reliability and overall functionality of the project.

How it is different

While many smart pumps in the market come with built-in remote monitoring systems, they are typically limited to their own brand and require full hardware replacement. InfuLink is different, it retrofits existing syringe pumps, regardless of brand, with intelligent remote monitoring capabilities. This means hospitals can upgrade their current equipment without costly replacements or vendor lock-in. InfuLink stands out by offering a universal, non-invasive solution that leverages computer vision and AI to monitor infusion states in real-time across various wards. Unlike most existing systems, InfuLink also integrates seamlessly with hospital EMR systems, enabling automatic infusion data logging and supporting more efficient, accurate clinical documentation. Its brand-agnostic design, real-time analytics, and EMR connectivity make it a truly unique and scalable approach to enhancing patient safety and workflow efficiency.

Future plans

To improve real-world usability, we plan to enhance the mechanical arm for broader pump compatibility and add a retractable cable mechanism. A compact custom PCB will reduce overall size and power consumption. On the software side, we aim to deploy local AI models to reduce dependence on cloud APIs. InfuLink features three main views: an overview for all wards, a ward-specific page, and a patient detail page. Integration with OpenEMR allows automatic updates to patient records, minimizing manual work and improving safety and efficiency in clinical settings.

Awards