AI_Malaria_Parasite_Detector

AI Malaria Detection System - Ghana Model

Overview

This system uses deep learning (MobileNetV2) to detect malaria parasites in blood smear images. The model has been fine-tuned specifically on Ghanaian blood smear samples to ensure accuracy with local staining and imaging conditions.

Performance Metrics

• Overall Accuracy: 97.2%
• Parasitized Detection: 97.8%
• Uninfected Rejection: 92.3%
• Model Size: 2.54 MB (mobile-optimized)

Folder Structure

/malaria_detector/
  ├── models/
  │   ├── local_finetuned_model.h5
  │   └── malaria_model.tflite
  ├── scripts/
  │   ├── predict_patches.py
  │   └── requirements.txt
  ├── evaluation/
  │   ├── final_report.pdf
  │   └── confusion_matrix.png
  └── README.md

Deployment & Usage

1. Clone the Repository: Download the entire malaria_detector folder.
2. Install Dependencies: Navigate to the scripts/ folder and install required libraries:

   pip install -r requirements.txt
   

3. Model Location: Ensure local_finetuned_model.h5 and malaria_model.tflite are in the models/ folder.
4. Run Prediction: Use the predict_patches.py script to analyze whole slide images:

   python scripts/predict_patches.py /path/to/your/image.jpg [confidence_threshold]
   

   • [confidence_threshold] is optional; defaults to 0.5 (50%). Recommended: 0.5.

Contact

For questions or further assistance, please contact the development team.