-
Project Logo
Smart Hospital Pillow - AI-Powered Patient Monitoring System
Inspiration
Healthcare workers are the unsung heroes of our society, yet they face an overwhelming challenge: managing multiple critically ill patients simultaneously while making split-second, life-or-death decisions. During our research, we discovered that nurses often spend precious minutes manually checking on patients, prioritizing care based on incomplete information, and struggling to identify which patient needs immediate attention.
What if a simple hospital pillow could change everything?
We envisioned a smart pillow equipped with non-invasive sensors that could continuously monitor patients and use AI to tell nurses exactly who needs help right now and what to do about it. This project was born from a desire to augment human expertise with intelligent automation, giving healthcare workers superpowers to save more lives.
How It Works (3 Steps)
1️⃣ Continuous Monitoring
Pillow sensors track: Temperature, Pressure, Movement, Heart Rate, Oxygen
- Non-invasive
- Real-time data collection
- Handles multiple readings for trend analysis
2️⃣ Intelligent Prioritization
Custom algorithm calculates:
- Weighted satisfaction score (0-100)
- Risk level (Critical → Stable)
- Priority ranking (who needs attention first)
- Edge case detection (catches hidden critical parameters)
$$\text{Priority} = (100 - \text{Score}) \times 0.6 \times \text{Risk Multiplier} + \text{Critical Penalty}$$
3️⃣ AI Clinical Summaries
Google Gemini generates:
- Diagnosis based on vital signs
- Nursing interventions (specific actions)
- Documentation reminders
- Executive dashboard summary for shift briefings
Real Example
Emma Rodriguez - ICU-101
Temperature: 41.5°C (Score: 5/100) ← CRITICAL
Pressure: 85 mmHg (Score: 15/100) ← CRITICAL
Overall: 13.05/100
Priority Score: 300.85 (Rank #1)
AI Summary:
"Critical sepsis. Immediate actions: 1) Notify physician, 2) Prepare emergency response, 3) Real-time monitoring, 4) Check patient NOW."
Tech Stack
- Backend: Node.js + Express + PostgreSQL
- AI: Google Gemini 2.5 Flash
- Algorithm: Multi-stage weighted scoring with edge case detection
- Performance: < 2 seconds per patient analysis
Impact
✅ Faster triage - Instant priority ranking
✅ Better outcomes - Edge case detection saves lives
✅ Time saved - Nurses focus on care, not data gathering
✅ Actionable intelligence - AI provides specific interventions
What Makes Us Different
| Feature | Traditional Monitoring | Our System |
|---|---|---|
| Prioritization | Manual/subjective | AI-powered ranking |
| Edge Cases | Often missed | Automatically detected |
| Actionability | Raw data dumps | Specific clinical actions |
| Intelligence | None | Gemini AI diagnosis |
The Math (Why It Works)
Weighted Average: $$S = \frac{\sum (\text{parameter} \times \text{weight})}{\sum \text{weights}}$$
Edge Case Override:
IF any_parameter < 30 THEN risk ≥ High Risk
Example: Patient with 4/5 normal vitals but oxygen = 25
→ Overall score = 75 (seems OK)
→ Our system: CRITICAL (respiratory failure detected)
Results
- 9 test patients validated
- 3 critical cases correctly prioritized
- 2 fluctuating conditions tracked across multiple readings
- 100% edge case detection accuracy
Log in or sign up for Devpost to join the conversation.