SCAnED - An Open-source Skin Segmentation Macro for Semi-automated Cell and Nuclei Detection in Epidermal and Dermal Skin Compartments

In our work, we investigate regulatory mechanisms that maintain tissue homeostasis at molecular, cellular, structural, and mechanical levels. Our goal is to understand pathological conditions, such as skin inflammation or cancer. With new high resolution imaging and multiomics tools, we can now map cellular and gene expression changes in tissues in 2D and 3D.

This is really pushing our understanding of how diseases develop and progress. One of the big challenges right now is the massive amount of data generated by advanced microscopy. And while there are more tools available to analyze images, they're often not user-friendly for people without a lot of experience, or are quite expensive.

Our protocol offers a free user-friendly tool for analyzing human skin cells in both epidermis and dermis. It improves accuracy over general tools, and doesn't require coding or prior image analysis experience. To begin, open the immunofluorescence images in Fiji or Image J, using bio formats with the default settings.

Do not split the channels during this step. Download the SCAnED macro and run it by selecting plugins. Then macros and clicking run.

Select the SCAnED macro script file when prompted. Follow the onscreen instructions provided by the macro. Select the E-cadherin channel when prompted by the macro.

Next, navigate to image, choose adjust, then threshold to manually adjust the threshold for the E-cadherin channel. Click OK once the epidermal region is fully captured. Then select the DAP-E channel when prompted to proceed with segmentation.

Auto define the intensity levels for nuclei segmentation when using the threshold function. For StarDist, open the plugin in Image J.Load the nuclei detection model into the StarDist graphical interface, and click on the nuclei channel to select it for analysis. Now, adjust the tile settings in StarDist to match the expected patch size of 256 by 256 pixels.

When prompted, select yes to measure marker intensity within the nuclei. Then repeat the segmentation and analysis process for each fluorescence channel. After all channels are processed, save the output data as comma separated value files as instructed by the macro.

For cytoplasmic region, when nuclei segmentation has been performed using the threshold function, apply binary dilation to expand the nuclear region of interest. When StarDist has been used for nuclei segmentation, apply enlargement expansion to the nuclear region of interest. Select yes when prompted to measure marker intensity within the whole cell region.

Then save the resulting data files as comma separated value format according to macro instructions. Establish marker specific intensity thresholds using isotype control samples. Download the appropriate Jupyter notebook, and navigate to Google Collab to select file, followed by open notebook.

Choose the file and run the code cells sequentially. When prompted with the choose files option, click the button and upload the comma separated value file. For each antibody, calculate the mean fluorescence intensity using the isotype control files.

To determine the cutoff thresholds, open the dot plot notebook, upload the isotype control data, and run the code. Based on the distribution of dots in the plots for both isotype control and specific antibody data, select an intensity threshold to distinguish background signal from true marker expression. Use this threshold to classify each cell as positive or negative for the given marker.

Finally, create dot plots to visualize co-expression patterns of markers by running the dot plot code in the Google Collab notebook. In manually segmented images, vimentin positive cells showed a strong cytoplasmic signal with clear ROI boundaries covering the full cell body. The SCAnED semi-automated segmentation captured vimentin positive cells with similar boundaries using a threshold and dilation approach.

Quantitative comparison of vimentin intensity in 21 cells showed no significant difference between manual and SCAnED segmentation. Histogram analysis revealed a higher frequency of vimentin positive cells in psoriasis dermis compared to healthy skin. Dot plot analysis of dermal cells showed increased CD-90 positive vimentin positive cells in psoriasis compared to healthy skin.

In the epidermis, psoriasis samples showed an increase in CD-90 positive vimentin negative cells, possibly keratinocytes, compared to healthy skin. Vimentin positive cell counts in psoriatic and healthy epidermis were comparable between SCAnED and QuPath. visual segmentation showed SCAnED more accurately assigned border cells to the epidermis than QuPath, which misassigned several.