Automated nailfold capillary density measurement method based on improved YOLOv5.

Area of Science:

• Biomedical Engineering
• Medical Imaging Analysis
• Physiology

Background:

• Nailfold capillary density is crucial for assessing nailfold health.
• Current clinical analysis of nailfold images is often subjective.
• Objective and rapid measurement methods are needed.

Purpose of the Study:

• To develop an automated method for measuring nailfold capillary density.
• To improve the accuracy and efficiency of nailfold image analysis.
• To provide an objective tool for clinicians.

Main Methods:

• An improved You Only Look Once version 5 (YOLOv5) algorithm was utilized.
• Novel techniques including 9mosaic and spatial pyramid pooling were incorporated.
• Distal capillaries were detected and filtered using the 90° method for density calculation.

Main Results:

• The proposed automated method achieved an Average Precision (AP)@0.5 of 85.2%.
• This represents significant improvements over original YOLOv5, YOLOv6, and R-CNN methods.
• Results demonstrated high consistency with expert manual calculations.

Conclusions:

• The improved YOLOv5 algorithm offers an effective and rapid solution for automated nailfold capillary density measurement.
• This automated approach enhances objectivity and accuracy in nailfold health assessment.
• The method shows superiority to existing automated techniques and manual expert analysis.