Stroke Analysis API Documentation

Overview

This API provides endpoints for stroke prediction analysis using different modalities: facial images, arm movement data, and speech audio. Each endpoint accepts specific file types and returns standardized prediction results.

Base URL

https://medaistroke.duckdns.org/api

Endpoints

1. Face Analysis

Analyzes facial images to predict stroke probability.

Endpoint: /face
Method: POST
Content-Type: multipart/form-data

Request

Parameter	Type	Required	Description
image	File	Yes	Image file containing a face

Response

{
    "message": "Face analysis completed",
    "result": {
        "stroke": 0,        // 0: No stroke, 1: Stroke
        "score": 0.873      // Probability score (0-1)
    }
}

2. Arm Movement Analysis

Analyzes arm movement data to predict stroke probability.

Endpoint: /arm
Method: POST
Content-Type: multipart/form-data

Request

Parameter	Type	Required	Description
csv	File	Yes	CSV file containing arm movement data

Response

{
    "message": "Arm analysis completed",
    "result": {
        "stroke": 0,        // 0: No stroke, 1: Stroke
        "score": 0.873      // Probability score (0-1)
    }
}

3. Speech Analysis

Analyzes speech audio to predict stroke probability.

Endpoint: /speech
Method: POST
Content-Type: multipart/form-data

Request

Parameter	Type	Required	Description
audio	File	Yes	Audio file containing speech

Response

{
    "message": "Speech analysis completed",
    "result": {
        "stroke": 0,        // 0: No stroke, 1: Stroke
        "score": 0.873      // Probability score (0-1)
    }
}

Error Responses

Bad Request (400)

Returned when the request is invalid (e.g., missing file, invalid file format)

{
    "error": "No image file"
}

Internal Server Error (500)

Returned when an unexpected error occurs during processing

{
    "error": "Internal Server Error",
    "message": "Error description"
}

Dependencies

• Python 11
• Flask 3.0.3
• TensorFlow 2.17.0
• OpenCV 4.10.0
• scikit-learn 1.5.2
• Additional dependencies listed in requirements.txt

Running the API

1. Install the required dependencies:

pip install -r requirements.txt

2. Ensure all model files are present in the configured TRAINED_MODELS_DIR:

   • face_model.pkl
   • arm_model.pkl
   • speech_model.keras

3. Start the Flask server:

python app.py

Notes

• All prediction endpoints return a standardized response format with a stroke prediction (0 or 1) and a confidence score
• In production mode, detailed error messages are suppressed for security
• Debug mode can be enabled for detailed error tracking during development