Image channel masking and segmentation

Image Channel Masking and Segmentation

License: restricted — no clear open-source license detected for the underlying tool; verify licensing before commercial use or redistribution. <!-- asb-license-banner -->

Summary

Apply pixel-level masking and segmentation operations to LA-ICP-MS elemental image arrays to isolate regions of interest and create binary or labeled masks. This skill enables selective extraction of chemical data by threshold, user-defined masks, k-means clustering, or Otsu's method, producing new derived images for downstream colocalization and quantitative analysis.

When to use

When you have multi-channel LA-ICP-MS images and need to isolate specific elemental regions (e.g., pixels above a threshold, clustered by k-means, or segmented by Otsu's method) to exclude background, select foreground, or partition spatial domains for independent analysis of colocalization or abundance.

When NOT to use

• Input image has already been binarized or manually segmented by non-automated means and you need only to apply it; use the direct mask operation instead of re-segmenting.
• Your goal is to measure colocalization across the entire image without spatial isolation; masking will discard data and may bias summary statistics.
• The elemental signal is uniformly distributed or lacks clear spatial structure; automatic thresholding or clustering will not resolve meaningful regions.

Inputs

• Multi-channel LA-ICP-MS image array (registered in pewpew open images)
• Threshold value (numeric scalar for threshold or Otsu operations)
• K parameter (integer count of clusters for k-means)
• Pre-computed mask image (for mask operation)

Outputs

• Binary or categorical mask image (NaN for excluded pixels, numeric for included or clustered)
• Registered masked image in pewpew open images (ready for further filtering, export, or colocalization calculation)

How to apply

Load the registered elemental image array into the Calculator or direct masking/segmentation module. Choose the appropriate filter step based on your segmentation goal: use threshold |image, value to create a binary mask by setting all pixels below a cutoff to NaN; use mask |image, mask_image to select pixels passing a pre-computed mask; use segment |image, threshold to generate thresholds-based segmentation; use kmeans |image, k to partition pixels into k clusters by spectral similarity; or use otsu |image to automatically compute an optimal intensity threshold. Evaluate the resulting mask or labeled array visually and quantitatively (e.g., region size, mean intensity within masked regions) to confirm that foreground/background separation or clustering is appropriate for your colocalization hypothesis.

Related tools

• pewpew (GUI for importing and interactive filtering of LA-ICP-MS images; hosts masking, segmentation, and threshold operations in the Calculator and direct filter UI.) — https://github.com/djdt/pewpew
• pewlib (Underlying Python library that implements image array I/O, arithmetic evaluation, and filter step execution (threshold, mask, segment, kmeans, otsu).) — https://github.com/djdt/pewlib

Evaluation signals

• Resulting mask image contains only NaN and numeric values (no negative or unintended values); spatial structure matches visual inspection of the input image.
• For threshold or Otsu: histogram or region-size statistics show bimodal separation; foreground region size is consistent with domain knowledge (e.g., cell or tissue area).
• For k-means: cluster labels are stable across repeated runs (deterministic seed or convergence check); cluster centroids are well-separated in intensity/elemental space.
• For segment: resulting mask can be overlaid on the original image with expected foreground–background spatial alignment.
• Pixels set to NaN in the mask are reproducibly excluded from downstream colocalization or intensity calculations; unmasked regions retain original intensity values.

Limitations

• Threshold and Otsu methods assume unimodal or bimodal intensity distributions; multimodal or skewed elemental distributions may produce spurious thresholds.
• K-means clustering requires pre-specification of k; no automatic k-selection is provided in the described workflow; results are sensitive to initialization.
• Segmentation and masking operations are applied independently to each image; no multi-image statistical thresholding or cross-channel spatial coherence constraints are mentioned.
• NaN pixels introduced by masking propagate through subsequent arithmetic and filtering operations; users must be aware that masked pixels will remain NaN in derived images unless explicitly handled.

Evidence

• [methods] mask |image, mask |selects from image using mask: "mask |image, mask |selects from image using mask"
• [methods] segment |image, threshold |creates a mask image of thesholds: "segment |image, threshold |creates a mask image of thesholds"
• [methods] kmeans |image, k |array of lower k-means bounds: "kmeans |image, k |array of lower k-means bounds"
• [methods] otsu |image |Otsu's method of image: "otsu |image |Otsu's method of image"
• [methods] threshold |image, value |sets all pxiels below value to nan: "threshold |image, value |sets all pxiels below value to nan"
• [other] Handle NaN propagation and conditionals: pixels failing thresholds become NaN; conditionals branch per-pixel.: "Handle NaN propagation and conditionals: pixels failing thresholds become NaN; conditionals branch per-pixel."
• [other] Pew² provides a Calculator module that evaluates per-pixel expressions over element channels to produce new derived image arrays, supporting conditional logic and arithmetic operations on elemental data.: "Pew² provides a Calculator module that evaluates per-pixel expressions over element channels to produce new derived image arrays, supporting conditional logic and arithmetic operations on elemental"
• [intro] Pew² is a GUI for importing and processing line-by-line, spot-wise and ablation-time-aligned LA-ICP-MS data: "Pew² is a GUI for importing and processing line-by-line, spot-wise and ablation-time-aligned LA-ICP-MS data"
• [readme] Pewlib is a library for importing, processing and exporting LA-ICP-MS data.: "Pewlib is a library for importing, processing and exporting LA-ICP-MS data."