An artificial intelligence-based approach for human parasite egg segmentation and classification

Intestinal parasitic infections are a significant global health issue, recognized by the World Health Organization (WHO) as a major cause of disease. Current diagnostic methods rely on labor-intensive manual examination of fecal samples under a microscope. This study aims to overcome these challenges by leveraging artificial intelligence (AI) to automate the identification of parasitic eggs in laboratory settings. To enhance image clarity and remove noise from microscopic fecal images, we employed the Block-Matching and 3D Filtering (BM3D) technique, which effectively addresses Gaussian, Salt and Pepper, Speckle, and Fog Noise. Contrast enhancement between subjects and the background was achieved using Contrast-Limited Adaptive Histogram Equalization (CLAHE). A U-Net model was utilized for image segmentation, followed by a watershed algorithm to extract Regions of Interest (ROI) from the segmented images. Finally, a Convolutional Neural Network (CNN) was developed for classification through automatic feature learning in the spatial domain. The U-Net model, optimized using the Adam optimizer, demonstrated excellent performance with 96.47% accuracy, 97.85% precision, and 98.05% sensitivity at the pixel level. At the object level, it achieved 96% Intersection over Union (IoU) and a 94% Dice Coefficient. The CNN classifier achieved 97.38% accuracy, with macro average F1 scores of 97.67%. This study presents an innovative AI-based approach for diagnosing intestinal parasitic infections. By integrating advanced image filtering, segmentation, and classification techniques, the proposed method shows promise in improving diagnostic efficiency and reducing reliance on manual processes. • Enhance image clarity and remove noise for precise parasite detection in health diagnostics. • Segment images effectively using an optimized artificial intelligence approach for improved health analysis. • Extract key features automatically to enhance classification accuracy in health imaging. • Achieve high precision in diagnosing parasitic infections with artificial intelligence for better health outcomes. • Reduce manual effort in laboratory diagnostics through automation.