X-Detect Machine Learning Models

X-Detect Machine Learning Models is a project that utilizes machine learning algorithms to predict diseases in the chest or thoracic area based on input images of X-ray scans. This project is designed for healthcare professionals, researchers, and developers working in the field of medical imaging analysis.

Endpoint /predict

This endpoint allows you to load a model and make predictions on an uploaded image file.

Dependencies

• Express.js
• TensorFlow.js (Node.js version)
• Multer
• Sharp
• Moment Timezone

Installation

1. Install the required dependencies by running the following command:

npm install express @tensorflow/tfjs-node multer sharp moment-timezone

2. Create a new file, e.g., index.js, and copy the provided code into it.

3. Run the server by executing the following command:

node index.js

4. The server will start running on port 8080. You can access the endpoint at http://localhost:8080/predict.

Usage

Request

• Method: POST
• Endpoint: /predict
• Content-Type: multipart/form-data

Request Parameters

• image: The image file to be processed. This should be included as a form-data field with the key image.

Example

• Key: image
• Value: Select File and choose your X-ray image file.

Response

The server will respond with a JSON object containing the predictions and additional information.

Response Format

• predictions: An object containing the labels as keys and the predicted percentages as values.
• maxLabel: The label with the highest predicted percentage.
• label: The label name.
• percentage: The percentage value.
• created: The current date and time in the Asia/Jakarta timezone.
• additionalInfo: Additional information based on the predicted label.
• description: Description of the condition.
• symptoms: Array of symptoms associated with the condition.
• nextSteps: Array of recommended next steps for the condition.
• recommendation: Recommendation for further action.
• action: Action to be taken.
• message: Message suggesting consultation with a specialist based on the predicted label.

Example Response

{
"predictions": {
 "Mass": "60.12%",
 "Nodule": "20.45%",
 "Normal": "5.80%",
 "Pneumonia": "7.33%",
 "Tuberculosis": "6.30%"
},
"maxLabel": {
 "label": "Mass",
 "percentage": "60.12%"
},
"created": "6/8/2023, 10:30:00 AM",
"additionalInfo": {
 "description": "Adanya massa atau tumor di paru-paru, yang bisa bersifat jinak atau ganas.",
 "symptoms": ["Batuk", "Sesak napas", "Nyeri dada"],
 "nextSteps": [
   "Konsultasikan dengan spesialis untuk evaluasi lebih lanjut",
   "Tambahan tes mungkin diperlukan",
   "Ikuti rencana perawatan yang direkomendasikan"
 ]
},
"recommendation": {
 "action": "Mohon segera lakukan konsultasi dengan dokter spesialis",
 "message": "Berdasarkan prediksi Mass, disarankan untuk konsultasikan dengan spesialis untuk evaluasi lebih lanjut dan pengobatan yang tepat."
}
}

Note: Ensure that the model file model.json is located in the 90acc+/ directory relative to the index.js file.

Please adjust the code and instructions as necessary for your specific use case.

Thankyou!