Automated Detection of Pinworm Parasite Eggs Using YOLO Convolutional Block Attention Module for Enhanced Microscopic Image Analysis

Esraa Hassan^1*Felwah Alqahtani^2*Samar Elbedwehy¹Amira Samy Talaat³

• ¹Faculty of Artificial Intelligence, Kafrelsheikh University, Kafr el-Sheikh, Egypt
• ²College of Computer Science, King Khalid University, 263, Abha 61471, Saudi Arabia, Abha, Saudi Arabia
• ³Computers and Systems Department, Electronics Research Institute, Cairo, 12622, Egypt, cairo, Egypt

Parasitic infections are a significant public health concern, particularly in healthcare and community settings where rapid and accurate diagnosis is critical for effective treatment and prevention. Traditional parasite detection methods are based on manual microscopic examinations, which can be time-consuming, labor-intensive, and prone to human error. However, recent advancements in microscopic imaging and automated detection technologies improve diagnostic accuracy and efficiency. Among these advancements, deep learning frameworks have emerged as powerful tools for enhancing medical parasitology diagnostics. These frameworks can analyze complex microscopic images, identify parasitic elements, and provide precise localization, supporting healthcare professionals in making informed decisions and using sophisticated algorithms. This study presents a robust architectural framework designed to automate the detection of pinworm parasite eggs in microscopic images. The proposed approach, the YOLO Convolutional Block Attention Module (YCBAM) architecture, integrates YOLO with self-attention mechanisms and the Convolutional Block Attention Module (CBAM). The proposed combination enhances the model's ability to identify and locate pinworm eggs even in challenging image backgrounds, significantly improving diagnostic accuracy. The experimental evaluation of the YCBAM architecture leads to better results, with a precision of 0.9971, a recall of 0.9934, and a training box reduced to 1.1410, indicating efficient learning and convergence. The model also obtained a mean Average Precision (mAP) of 0.9950 at an IoU threshold of 0.50, and a mAP50-95 score of 0.6531 across varying IoU thresholds.