TriaSense--Smart Decisions, Shared Humanity

Overview

In this project, we will prototype a frontline rural clinic triage product: TriaSense.

Learning Objectives

• Get familiar with buttons
• Implement a classification system using decision tree
• Create an interactive and engaging user experience

Hardware Requirements

• ESP32 development board
• 4 push buttons
• OLED display (Adafruit SSD1306, 128x64)
• USB cable
• Breadboard and jumper wires

Software Requirements

• Arduino IDE or PlatformIO
• ESP32 board support package
• Required libraries:
  • EloquentTinyML
  • Adafruit SSD1306
• Dependencies defined in requirements.txt

Project Structure

.
├── src/                    # Source code directory for laptop
│   └── TriaSense_laptop.py    # Python script for training the model
    ├── requirements.txt       # Python dependencies
├── TriaSense_ESP32/
    └── TriaSense.ino   # ESP32 Arduino sketch
├── assets/                 # Assets directory
│   └── TriaSense.png   # Wiring diagram
└── README.md               # Lab manual

Dataset Structure

The dataset is stored in a CSV file with the following structure: (Note: Revise the file name accordingly for your setup)

Question 1	Question 2	Question 3	...	Decision

Label column: TriaSenseLabel

Valid labels: BLUE_Observe, GREEN_Routine, YELLOW_Priority, RED_Urgent

Features

• Each question column contains integer values (0-3) representing the user's response
• The TriaSenseLabel column contains the target variable (BLUE_Observe, GREEN_Routine,YELLOW_Priority, RED_Urgent)
• Each row represents a patient record

Tasks

1. Navigate to src directory. Create a virtual environment, activate it, and install the required libraries.

   For MacOS:

   python3 -m venv .venv
   
   source .venv/bin/activate
   
   pip install -r requirements.txt

   Fow Windows:

   python -m venv .venv
   
   .venv/Scripts/Activate
   
   pip install -r requirements.txt

2. Use src/TriaSense_laptop.py to train your decision tree and convert it to C++ for ESP32.

3. Wire your ESP32 with buttons and OLED display. An example is shown in the figure below. Update the sketch TriaSense.ino accordingly, depending on how you wire the TriaSense.

Discussion

• Play with TriaSense. Sketch the workflow of TriaSense.
• What additional sensors or hardware could enhance the user experience?
  • Does decision tree remain suitable for your choice of new sensors? If yes, carefully justify your answer. If not, what ML model would you use and explain why.
• How would you pitch TriaSense to a rural hospital administrator?
• What are potential privacy or data ownership concerns when recording triage data in rural contexts?
• How do technical constraints, user needs, and business realities interact in shaping TriaSense’s design?
• What cultural or social biases might emerge in the question phrasing or decision rules?
• How do technical constraints, user needs, and business realities interact in shaping TriaSense’s design?