TL;DR
Researchers have developed a new ImageJ plugin called "Wound Healing Size Tool" that automates the analysis of in vitro scratch wound healing assays. This tool significantly speeds up the process, improves accuracy, and accounts for variations like scratch inclination, making it a valuable asset for studying cell migration and proliferation.
What Problem Does This Paper Address?
The in vitro scratch wound healing assay is a widely used, cost-effective method for studying cellular migration and proliferation, crucial processes in wound healing and disease progression like cancer. While ImageJ is a popular open-source tool for analyzing these assays, existing methods often require extensive manual parameter tuning. This manual process is time-consuming, limits the number of samples that can be analyzed (throughput), and can introduce user variability, especially for inexperienced researchers. The paper aims to overcome these limitations by providing an automated, user-friendly plugin for ImageJ.
Methods
The core of this work is the development of a new ImageJ plugin, the "Wound Healing Size Tool." The plugin employs a classic computer vision segmentation algorithm that focuses on analyzing the variance of neighboring pixel intensities. The rationale is that open wound areas have more uniform pixel intensities compared to cell monolayers, which exhibit higher intensity variation due to cell presence.
The process involves:
- Contrast Enhancement: To increase the intensity variance within the cell monolayer.
- Variance Filter: Applied to calculate intensity variance in the neighborhood of each pixel, creating an image where cell areas have high intensity and wound areas have low intensity.
- Binarization: A simple thresholding step is applied to the processed image to create binary segments representing the wound area.
- Morphological Reconstruction (Hole Filling): To include small regions within the wound area that might have been missed (e.g., cell islets).
- Largest Connected Component Selection: To eliminate any falsely detected wound regions within the cell monolayer.
A key innovation is the plugin's ability to correct for the inclination of the scratch using trigonometric relations, ensuring more accurate width measurements. It can analyze individual images or image stacks and is compatible with images from bright field, phase contrast, and fluorescence microscopy.
Datasets
The paper describes experiments using HaCaT cells (human skin keratinocytes) exposed to conditioned medium from human adipose-derived mesenchymal stem cells (hAdMSCs). The evaluation of the plugin involved analyzing 30 images for comparison with other methods and 60 wound images for performance assessment. No new benchmark datasets were introduced; rather, existing experimental data was used for validation.
Key Insights and Ideas
- Variance as a Key Feature: The paper highlights that analyzing pixel intensity variance is a robust way to distinguish between cell-covered areas and open wound spaces in microscopy images.
- Inclination Matters: Wound width measurements can be significantly affected by the angle of the scratch. The plugin's ability to correct for this inclination is a crucial improvement over methods that assume a perfectly vertical scratch.
- Automation is Key: The development of an automated plugin addresses the bottleneck of manual analysis, making high-throughput screening of wound healing assays feasible.
- Versatility: The plugin's compatibility with different microscopy types (bright field, phase contrast, fluorescence) and its ability to handle both individual images and time-lapse stacks make it broadly applicable.
Key Results
- The "Wound Healing Size Tool" plugin was developed and tested.
- It quantifies wound area, wound coverage of total area, average wound width, and standard deviation of wound width.
- The plugin demonstrated a reduction in scratch width variability of approximately 50% when compared to manually made scratches using a pipette tip, thanks to a custom-designed mold.
- The plugin's analysis time for 60 images was 5-8 minutes, significantly less than manual methods.
- When compared to other ImageJ plugins (MRI, MiToBo) and manual measurements, the plugin showed no significant differences in area measurement for regularly shaped scratches. For irregularly shaped scratches, it provided results closer to MiToBo.
- The plugin's average width measurements were comparable to manual methods, though the standard deviation showed a significant difference (likely due to the need for more measurements for accurate standard deviation estimation).
- The plugin's inclination correction for width measurement resulted in an average width difference of only 1.35 ± 0.59%.
Limitations and Future Work
- Parameter Tuning: While automated, the plugin still requires user-defined input values for parameters like variance window radius, threshold value, and saturation percentage. The authors suggest testing these parameters on a few images with different combinations to find optimal settings for a given dataset.
- Standard Deviation Estimation: The paper notes a significant difference in standard deviation of width compared to manual methods, suggesting that a larger number of width measurements (at least 100) might be needed for a more accurate standard deviation estimate.
- Future Work: The paper doesn't explicitly detail future work, but the successful validation across different imaging modalities and the comparison with existing tools suggest potential for wider adoption and integration into research workflows. The open-source nature of the plugin also invites community contributions and improvements.
Summary updated