An automated malaria cells detection from thin blood smear images using deep learning

Area of Science:

• Medical diagnostics
• Artificial intelligence in healthcare
• Parasitology

Background:

• Accurate malaria diagnosis is vital for effective treatment.
• Microscopic examination of blood films is the current standard but requires skilled personnel.
• Deep learning offers potential for automated malaria detection.

Purpose of the Study:

• To identify the optimal deep learning object detection architecture for malaria diagnosis.
• To evaluate the performance of Yolov4, Faster R-CNN, and SSD 300 models.
• To assess the generalizability of the best-performing model on independent datasets.

Main Methods:

• Trained object detection models (Yolov4, Faster R-CNN, SSD 300) on malaria-infected blood cell images.
• Utilized an 80/20 train-test split across all five malaria parasite species and four infection stages.
• Evaluated model performance using precision, recall, F1-score, and mean average precision (mAP).

Main Results:

• Yolov4 achieved high performance: 83% precision, 95% recall, 89% F1-score, and 93.87% mAP at a 0.5 threshold.
• The Yolov4 model demonstrated strong generalization capabilities on an independent dataset.
• Object detection models eliminate the need for single-cell image training.

Conclusions:

• Yolov4 is a suitable deep learning alternative for detecting malaria-infected cells in whole blood smear images.
• Object detection complements deep learning classification by simplifying the detection process.
• AI-driven object detection presents a feasible approach for malaria diagnosis in diverse settings.