Research on Blood Cell Image Detection Method Based on Fourier Ptychographic Microscopy

Area of Science:

• Pathology
• Medical Imaging
• Computer Vision

Background:

• Autonomous Fourier Ptychographic Microscopy (FPM) enables high-resolution, large field-of-view imaging crucial for detailed pathological observations.
• Red blood cells are vital biomarkers for assessing oxygen transport and guiding clinical decisions.

Purpose of the Study:

• To develop an advanced algorithm for accurate and robust detection of red blood cells using FPM.
• To enhance the generalization and robustness of blood cell detection models through data preprocessing and architectural innovations.

Main Methods:

• Construction of a blood cell dataset utilizing an FPM experimental platform.
• Implementation of four image enhancement strategies to improve model generalization.
• Proposal of the SCD-YOLOv7 algorithm featuring C-MP and DELAN modules for optimized feature extraction, a Sim-Head detection head for improved small target detection, and a Focal-EIoU loss function for enhanced accuracy and convergence.

Main Results:

• The SCD-YOLOv7 algorithm achieved a blood cell detection accuracy of 92.4% on the FPM dataset, an improvement of 7.2%.
• The algorithm demonstrated a reduction in computational load by 14.6 G.
• Ablation and comparative experiments validated the effectiveness of the proposed modules and loss function.

Conclusions:

• The SCD-YOLOv7 algorithm significantly advances automated blood cell detection in pathology.
• The integration of FPM imaging with sophisticated deep learning models offers a powerful tool for clinical diagnostics.
• The developed method shows potential for improving the efficiency and accuracy of hematological analyses.