Pixie AI

Inspiration

Nurses in hospitals manage multiple patients daily and overwhelming workloads in high-stress environments. They often have to care for more patients than recommended, which leads to increased medical errors, burnout, and even patient mortality. For instance, a single additional patient per nurse can raise the likelihood of patient mortality by 7% and significantly increase nurse burnout rates (Penn LDI).

What it does

𝗣𝗶𝘅𝗶𝗲 𝗔𝗜 directly tackles this urgent issue. [Note: this tool is not for triage; it’s designed for patients who have been hospitalized for several days.]

Our innovative solution revolutionizes how nurses prioritize patient care. By integrating seamlessly with hospital systems, Pixie AI reduces the manual strain on nurses through a proactive approach:

𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗖𝗵𝗲𝗰𝗸-𝗜𝗻𝘀: Before each check-in, Pixie AI contacts patients via bedside intercoms, posing standardized questions. We made sure the AI sounds as natural and fast as possible. These interactions are recorded and analyzed for emotional and health cues (patient's mood, condition, note, etc), helping to prioritize care needs.
𝗦𝗺𝗮𝗿𝘁 𝗥𝗮𝗻𝗸𝗶𝗻𝗴 𝗦𝘆𝘀𝘁𝗲𝗺: Using tone analysis & sentiment analysis for the responses, our system evaluates the urgency of each patient's situation, ranking them to ensure nurses attend to the most critical cases first.
𝗥𝗲𝗮𝗹-𝗧𝗶𝗺𝗲 𝗨𝗽𝗱𝗮𝘁𝗲𝘀: Nurses and doctors can check the order of their visit anytime. We prioritize patients who need urgent help first. Nurses receive pushed alerts when there are emergencies via emails & SMS.
PDF View: Doctors and Nurses can view patient's history with their PDF medical records.
Similar Case Search: By typing the patient's ID, nurses can search for similar cases

Product Walkthrough

Data Collection: The process begins with Pixie AI’s integration into the hospital’s infrastructure, including bedside intercoms and EHR systems. Patients provide answers to standardized questions through bedside phones.
Sentiment and Tone Analysis: We use Google Gemini and Hume to analyze the sentiment and tone of these responses to evaluate the urgency of the patient’s condition.
Data Structuring:
Unstructured.io processes various documents (PDFs, check-in forms, research papers) and converts them into structured data.
Structured data is stored in MongoDB, utilizing vector search for efficient retrieval and analysis.
Dashboard and Decision-Making:
Ranking Patients: Based on analysis, Pixie AI ranks patients according to the urgency of their conditions. This ranking is displayed on the dashboard.
Similarity Search: MongoDB’s vector search finds similar cases based on patient data, providing insights and historical context

How to use it

Setting Up:

Clone this repository
Create environment settings: cp .env.example .env
Fill in the required environment variables in the .env file. Make sure you have all the required keys and secrets.

Instructions:

To run chat, go to backend , click on run.sh
In the backend folder, run the Streamlit UI: make run
All PDF records will be saved in records folder.
All QR code will be saved in qr_code folder.

How we built it

Pixie AI was built with the goal of optimizing nurses' workflow in hospitals. We wouldn't make it without Unstructured.io & MongoDB Atlas Vector Search.

As a POC, we want to keep it simple for both users and developers. We use HumeAI to build the chat interface faster. The patients' information, conversations, and records are saved in MongoDB clusters.

Lastly, we put a lot of effort in UX design to make sure doctors, patients, and hospitals can use it without knowing much about programming.

What we learned

How to use Unstructured.io & MongoDB

What's next for Pixie AI

Looking ahead, we plan to expand Pixie AI’s capabilities to include more languages and dialects, making it accessible to a broader range of healthcare facilities worldwide. We are also exploring additional AI features, such as predictive analytics, to anticipate patient needs before they arise. Moreover, ongoing partnerships with healthcare providers will be crucial for continuous improvement and wider adoption of Pixie AI. At the core, Pixie AI is about saving medical processes time, and let doctors have more freedom to care for their patients.