Plasmo3Net: A Convolutional Neural Network-Based Algorithm for Detecting Malaria Parasites in Thin Blood Smear Images

Live Demo

Article

Authors

Afolabi Owoloye^1,2,3 Funmilayo Ligali^1,3 Ojochenemi Enejoh⁴ Oluwafemi Agosile¹ Adesola Musa³ Oluwagbemiga Aina³ Adetunji Adewole⁵ Taiwo Idowu² Kolapo Oyebola^1,3

¹Centre for Genomic Research in Biomedicine (CeGRIB), College of Basic and Applied Sciences, Mountain Top University, Ibafo, Nigeria.
² Parasitology and Bioinformatics Unit, Department of Zoology, Faculty of Science, University of Lagos, Lagos, Nigeria.
³ Nigerian Institute of Medical Research, Lagos, Nigeria.
⁴ Genetics, Genomics and Bioinformatics Department, National Biotechnology Research and Development Agency, Abuja, Nigeria.
⁵ Computer Science Department, Faculty of Science, University of Lagos, Lagos, Nigeria

Figure 1.0: Experimental workflow for developing the convolutional neural network

🦟 Plasmo3Net: Malaria Detection Web App

A deep learning-powered web application for automated detection of Plasmodium falciparum in red blood cells (RBCs) using convolutional neural networks (CNNs). Achieves 99.3% accuracy in classifying infected vs. uninfected cells.

🔍 Features

• Cell Segmentation: Extracts individual RBCs from microscope slide images using OpenCV.
• CNN Classification:
  • Custom Plasmo3Net model (13-layer CNN) with 99.3% accuracy.
  • Benchmarked against ResNet50, InceptionV3, VGG16, and AlexNet.
• User-Friendly Interface:
  • Upload microscope slides or pre-segmented RBCs.
  • Real-time predictions with visual results.
  • Download segmented cells as a ZIP file.

📊 Performance Metrics

Metric	Plasmo3Net	ResNet50 (Baseline)
Accuracy	99.3%	97.9%
Precision	99.1%	97.6%
Recall	99.6%	98.3%
F1 Score	99.3%	97.9%

🚀 Quick Start

Prerequisites

• Python 3.8+
• TensorFlow 2.x
• Streamlit
• OpenCV

### Installation
1. Clone the repository:
   ```bash
   git clone https://github.com/CeGRIB/plasmo3net.git 
   cd plasmo3net

2. Install dependencies:

   pip install -r requirements.txt

3. Run the app:

   streamlit run app.py

4. Access the app at http://localhost:8501.

🧩 Workflow

1. Upload: Microscope slide image or segmented RBCs.
2. Segment: Automated RBC extraction (if using slides).
3. Classify: Predicts "Parasitized" or "Uninfected" for each cell.
4. Results: Displays predictions with confidence metrics.

📂 File Structure

.
├── app.py                # Streamlit application
├── model.h5              # Trained Plasmo3Net model
├── requirements.txt      # Dependencies
├── script/               # python scripts
├── RBC/                  # Processed RBC images
└── logo/                 # App assets

📄 Citation

If you use this work, please cite:

@article{
  authors={Owoloye et al.},
  title={Plasmo3Net: A Convolutional Neural Network-Based Algorithm for Detecting Malaria Parasites in Thin Blood Smear Images},
  journal={bioRxiv},
  year={2024},
  doi={10.1101/2024.12.12.628235v1}
}

🤝 Contribute

• Report bugs via Issues.
• Fork and submit Pull Requests.

📧 Contact

CeGRIB - @cegrib LinkedIn | Twitter