Deep Learning Based Automatic Malaria Parasite Detection from Blood Smear and its Smartphone Based Application

Area of Science:

• Medical diagnostics
• Artificial intelligence in healthcare
• Parasitology

Background:

• Malaria diagnosis relies on manual microscopy of blood smears by trained experts.
• This manual process is time-consuming, resource-intensive, and prone to human error.
• There is a critical need for automated, accurate, and efficient diagnostic tools to combat malaria.

Purpose of the Study:

• To develop and validate an automated Convolutional Neural Network (CNN) based model for malaria parasite detection in microscopic blood smear images.
• To optimize the model's accuracy and inference performance using techniques like knowledge distillation, data augmentation, and autoencoders.
• To assess the practical deployability and efficiency of the developed model in real-world scenarios.

Main Methods:

• An automated malaria diagnosis model was designed using Convolutional Neural Networks (CNNs).
• Techniques such as knowledge distillation, data augmentation, and autoencoders were employed for model optimization.
• Feature extraction was performed by a CNN, with classification using Support Vector Machines (SVM) or K-Nearest Neighbors (KNN) under three distinct training procedures.
• The model was evaluated for accuracy and computational efficiency (floating point operations).

Main Results:

• The developed deep learning model achieved a high diagnostic accuracy of 99.23% for detecting malarial parasites.
• The model requires minimal computational resources, with just over 4600 floating point operations.
• The miniaturized model demonstrated efficient performance, with inference times under 1 second per sample when deployed on mobile phones and a web application.

Conclusions:

• The automated CNN-based model offers a highly accurate and efficient solution for malaria diagnosis from microscopic images.
• The model's low computational requirements and fast inference times make it suitable for practical deployment in resource-limited settings.
• This AI-driven approach has the potential to significantly reduce the reliance on manual microscopy, improving malaria detection accessibility and speed.