Enhanced Detection of Leishmania Parasites in Microscopic Images Using Machine Learning Models.

Area of Science:

• Parasitology
• Medical Imaging
• Machine Learning

Background:

• Cutaneous leishmaniasis diagnosis is challenging due to variable results and operator dependence.
• Direct smear microscopy is a common diagnostic method but requires expertise.
• Developing automated systems can improve diagnostic accuracy and consistency.

Purpose of the Study:

• To develop and evaluate a machine learning-based system for detecting Leishmania spp. in direct smear microscopy images.
• To enhance the accuracy and objectivity of cutaneous leishmaniasis diagnosis.
• To investigate the effectiveness of image processing and machine learning for parasite identification.

Main Methods:

• Acquisition and labeling of 500 microscopy images.
• Image preprocessing and segmentation using Otsu, local thresholding, and IGMS.
• Extraction of phenotypic features (morphology, texture, color).
• Application and optimization of Artificial Neural Network (ANN), Support Vector Machine (SVM), and Random Forest (RF) models using Grid Search.

Main Results:

• The Support Vector Machine (SVM) model demonstrated superior performance.
• Achieved a sensitivity of 91.87% and a specificity of 89.21% at the crop level.
• Results show consistency and relevance compared to previous studies.

Conclusions:

• Machine learning techniques show significant potential for improving cutaneous leishmaniasis diagnosis.
• The developed SVM model offers a promising, accurate, and consistent tool for parasite detection.
• Automated analysis of microscopy images can overcome diagnostic challenges in leishmaniasis.