Methodology for comprehensive cell-level analysis of wound healing experiments using deep learning in MATLAB

Area of Science:

• Cardiovascular research
• Medical image analysis
• Cell biology

Background:

• Endothelial healing is critical for cardiovascular device outcomes.
• Accurate prediction of endothelial growth post-injury is essential.
• Current in vitro cell migration assays have limitations in image analysis accuracy and user dependence.

Purpose of the Study:

• To present an intelligent cell detection (iCD) approach for comprehensive analysis of in vitro wound healing assays.
• To enable precise prediction of endothelial growth and cell migration characteristics.
• To leverage deep learning for improved medical image analysis in cell-based assays.

Main Methods:

• Development of an intelligent cell detection (iCD) algorithm based on deep learning.
• Comparison of iCD with conventional analysis methods (adaptive threshold, Cell Image Velocimetry, Canny edge detection) in an in vitro wound healing assay.
• Analysis of cell density, cell velocity, monolayer movement, and gap closure at both cell and population scales.

Main Results:

• iCD demonstrated superior robustness against image distortion compared to adaptive threshold methods for cell density analysis.
• iCD cell velocity measurements agreed with manual detection, while Cell Image Velocimetry (CIV) underestimated velocity by 50%.
• iCD analysis of monolayer movement and edge protrusion showed comparable or superior accuracy to manual and conventional methods, with significantly lower error rates.

Conclusions:

• Deep learning-based iCD outperforms conventional methods for in vitro wound healing analysis.
• iCD provides timesaving, high-quality analysis at both cell and population scales.
• This approach enhances understanding of endothelial cell movement for research applications.