Parameter-Reduced YOLOv8n with GhostConv and C3Ghost for Automated Blood Cell Detection.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Hematology

Background:

• Accurate detection of blood cells is vital for automated hematological analysis and clinical diagnosis.
• Current methods may face challenges in efficiency and computational load for real-time applications.

Purpose of the Study:

• To develop an improved, lightweight YOLOv8n-based model for efficient and precise detection of red blood cells (RBCs), white blood cells (WBCs), and platelets.
• To evaluate the impact of GhostConv and C3Ghost modules on model performance and computational efficiency.

Main Methods:

• Utilized the YOLOv8n framework as a baseline for blood cell detection.
• Integrated GhostConv and C3Ghost modules to enhance model complexity and performance.
• Conducted ablation experiments on the BCCD dataset to assess module contributions.

Main Results:

• The baseline YOLOv8n achieved an mAP@0.5 of 0.9043.
• Incorporating GhostConv reduced parameters to 2.73 M with mAP@0.5 = 0.9040.
• The combined GhostConv and C3Ghost model achieved 0.9001 mAP@0.5 with only 1.71 M parameters.

Conclusions:

• The improved YOLOv8n model effectively balances high detection accuracy with reduced computational requirements.
• The lightweight framework is suitable for real-time blood cell analysis and integration into automated systems.
• This approach facilitates rapid and intelligent medical diagnostics in hematology.