The Artifact Detection extension uses computer vision to detect common artifacts in your image, such as blur, tissue folds, dust, imaging artifacts, etc. This tool will generate a cell annotation that will mark which cells are artifact, and which are true, non-artifact cells in the tissue. This enables downstream analysis on true cells.

The underlying artifact detection model is trained in a supervised fashion on a set of DAPI images with various artifacts.

Configuring new runs

To configure a new run, specify

• Title for the resulting annotation set
• [Optional] Description for the resulting annotation set
• Method used to run the model
• Regions to run the artifact detection on
• Segmentation Version for which cells to create the annotation with
• Nuclear Biomarker to use to calculate where the artifact is

Notes:

• This extension has been trained on DAPI. Results may vary for other biomarkers. Other high quality nuclear markers should retain similar performance.
• This extension may fail for particularly large images. We currently support sizes up to 50K x 50K pixels, and hope to expand this further shortly!
• Why aren’t all of my regions / segmentation versions appearing?
• What is the difference between the two methods?

Viewing results

Once a run is complete, you should be able to select that run within the extension to open the results in the Visualizer.

Alternatively, you can go directly to the Visualizer to open the resulting annotation and view it.

The image above shows an example with true cells in white and artifact cells in red.

If you have Workbench, you will additionally be able to leverage this annotation set in downstream analysis, to ensure you are analyzing only true, non-artifact cells. You can also use Workbench to push this annotation set back to the web portal as a quality control filter using the function upload_qc (demonstrated in this SpatialMap tutorial). We are anticipating further improvements to our web portal to allow you to filter out artifacts directly within the web view! Stay tuned for more.

Return to Home Directory