An automated neural network-based stage-specific malaria detection software using dimension reduction: The malaria...

Area of Science:

• Medical diagnostics
• Parasitology
• Computational biology

Background:

• Malaria cases increased 2019-2020 due to climate change and COVID-19.
• Artificial intelligence (AI) shows potential for improving malaria diagnosis.
• Current AI methods struggle with large parameter counts and small datasets in red blood cell (RBC) classification.

Purpose of the Study:

• To enhance malaria diagnosis performance by addressing limitations in AI-based RBC classification.
• To develop a tool for fast, high-accuracy malaria stage recognition.

Main Methods:

• Developed a Malaria Stage Classifier tool.
• Reduced input data dimensionality and used data augmentation for neural network (NN) training.
• Extracted one-dimensional cross-sections from individual RBC images for classification.
• Applied the method to light, atomic force, and fluorescence microscopy images.

Main Results:

• The Malaria Stage Classifier achieved fast, high-accuracy recognition of malaria blood stages.
• The method demonstrated applicability across various microscopy techniques.
• The approach effectively handles challenges of small training sets and uneven data distribution.

Conclusions:

• The developed AI tool offers a significant improvement for malaria diagnosis.
• The Malaria Stage Classifier is a versatile and efficient solution for identifying malaria in RBCs.
• This approach can aid in reversing the increasing trend of malaria cases and deaths.