Deep learning method for malaria parasite evaluation from microscopic blood smear.

Area of Science:

• Medical diagnostics
• Parasitology
• Computer science

Background:

• Malaria causes significant global morbidity and mortality, with approximately 597,000 deaths reported in 2024.
• Automated analysis of Plasmodium parasites in blood smears offers potential to enhance diagnostic accuracy and efficiency.

Purpose of the Study:

• To systematically review current methodologies for automated malaria diagnostics.
• To examine computer-assisted methods, databases, staining techniques, and diagnostic models.
• To identify limitations and contributions of recent studies in automated malaria detection.

Main Methods:

• Systematic literature review adhering to PRISMA guidelines.
• Searched Web of Science and Scopus for peer-reviewed studies from 2020-2024.
• Included studies using deep learning and machine learning for automated malaria detection from blood smears.

Main Results:

• The NIH database is a standardized resource for malaria diagnostics.
• Giemsa-stained thin blood smears are optimal for Plasmodium lifecycle observation.
• Identified top-performing models: ResNet/VGG (99.12% accuracy), popular custom CNNs (58% of studies), and CADx models.

Conclusions:

• Automated malaria diagnostics hold significant potential for reducing errors and improving accuracy.
• Deep learning models show high performance, yet data standardization and real-world implementation require further attention.
• Addressing current challenges can lead to more reliable and scalable diagnostic tools for global malaria control efforts.