Inspiration /
Our inspiration comes from the heart-breaking statistic about, unfortunately, the number of deaths in emergency rooms per year in Quebec and Canadian hospitals. Our project is mainly based on the attempt to reduce this number increasingly higher by introducing an AI-based hospital dynamic triage system. It allows a better priorization management in emergency rooms.
What it does
/ This AI-based Hospital Dynamic Triage System consists of a facial recognition system that analyzes the emotions, such as fear, sadness, happiness, or just a neutral patient, of a patient which combines the initial evaluation of the patient, such as his name, age, symptoms, conditions, and his arrival time at the emergency room. All this information is connected to DeepSeek API that will ensure the final analysis of the report, telling whether the patient needs a medical attention immediately to save his life, or not and classifying the patient into three stages of priority (somewhat similar as the initial triage in emergency rooms).
How we built it
/ There are two main parts of the project : the facial recognition system and the integration of DeepSeek API to ensure the final report of the patient. For the facial recognition system, we use an open source project that already does the analysis of the emotions of the patients (fear, sadness, neutral, happiness). We incorporated it with many fictional patients with a name, an age, symptoms, chronic conditions, and its arrival time. Through this data collection, we have built complete patient profiles that encompass not only their medical conditions but also their emotional states. We then make use of DeepSeek API that will write, based on all the data collected, the priority level of the patient. There are three levels, where the first one needs medical attention immediately, and the third is able to wait.
Challenges we ran into
/ The main challenge was about the facial recognition system. Facial recognition systems can struggle with variations in lighting, pose, facial expressions, and occlusions (e.g., masks, glasses, or hair covering the face). Also, integrating facial recognition with our existing triage system can be complex. So, we needed to test during a lot of time to ensure a good accuracy.
Accomplishments that we're proud of
/ First of all, we successfully integrated an AI-based system (DeepSeek API) to analyze patient data and prioritize cases based on symptoms, conditions, and emotional state, by creating a comprehensive system to manage all this patient data. We then successfully implemented a facial recognition system for patient identification and emotion detection, which was the main part of the project. Our main accomplishment remains in the introduction of an innovative solution to address a challenge in healthcare.
What we learned
/ We learned that technology can have a significant impact on real-world problems, such as improving the efficiency and accuracy of medical triage. Also, technical challenges, such as integrating facial recognition and optimizing performance, require creative problem-solving and persistence. Finally, integrating third-party APIs (DeepSeek) can be complex, especially when dealing with authentication, rate limits, and error handling.
What's next for AI-based Hospital Triage System
/ Firstly, we want to enhance our AI model, by improving the accuracy and reliability of the AI models used for triage and emotion detection (incorporating multimodal data). We would like to enable real-time integration with hospital systems for up-to-date patient information. Also, we want to improve the user experience, by making the system more intuitive and user-friendly for healthcare providers. It is also important to enhance the facial recognition system for better accuracy and security. Finally, we want to ensure the system is fair, transparent, and free from bias to ensure ethical integrity in our AI-based Hospital Dynamic Triage System.
Log in or sign up for Devpost to join the conversation.