Parasite.ai – An Automated Parasitic Egg Detection Model from Microscopic Images of Fecal Smears using Deep Learning Techniques
Sakthi Jaya Sundar Rajasekar, Gaurav Jaswal, Varalakshmi Perumal, Sasana Ravi, Varun Dutt · 2023
Intestinal parasitic infestations are one of the leading reasons of morbidity worldwide and has been recognized as one of the most significant causes of diseases by World Health Organization (WHO). The mainstay of diagnosis is through direct manual examination of fecal smears using microscope in the laboratory. However, there are several shortcomings in the manual parasite egg detection like its time-consuming nature, low sensitivity and mandatory requirement of skilled technicians. Utilization of Artificial Intelligence in automation of the laboratory identification of parasite eggs has been proposed in this work. Various Deep Learning models like Yolov8, Detectron2, Inception v3 and YOLOs have been utilized for parasite egg detection task and their performance has been evaluated. Our results indicate that YOLOv8 with SGD optimizer has yielded superior performance, mAP of 0.92 and F1-score of 98%, compared to the other models evaluated and the existing literature. Its high accuracy and fast convergence make it an ideal choice for the parasitic egg detection tasks.