Robust Image Processing Framework for Intelligent Multi-Stage Malaria Parasite Recognition of Thick and Thin Smear...

Area of Science:

• Medical diagnostics
• Computational biology
• Parasitology

Background:

• Malaria diagnosis relies on manual microscopy, which is labor-intensive and prone to inconsistencies.
• Automated methods are needed to improve the accuracy and efficiency of malaria parasite detection and staging.

Purpose of the Study:

• To develop a standardized image processing and machine learning framework for malaria parasite segmentation, species recognition, and staging.
• To address the limitations of manual microscopic examination in malaria diagnosis.

Main Methods:

• A segmentation framework using Phansalkar thresholding and enhanced k-means (EKM) clustering was developed.
• A multi-stage classifier, including a random forest (RF) model, was designed for parasite detection, species recognition, and staging.
• The framework was applied to segment parasite stages of *P. falciparum* and *P. vivax*.

Main Results:

• Phansalkar thresholding achieved 99.86% accuracy in segmenting thick smear images.
• EKM clustering demonstrated high accuracy (99.20%) and F1-score (0.9033) in segmenting malaria stages.
• The RF classifier attained accuracies of 86.89% for detection, 98.82% for species recognition, and 90.78% for staging.

Conclusions:

• The proposed framework offers a versatile and accurate automated solution for malaria parasite detection and staging.
• This approach can potentially be extended to other malaria species, improving diagnostic capabilities in endemic regions.