Review the mask before accepting measurements
Use the segmentation overlay, ROI controls and accepted/rejected object review to check whether the mask follows real object boundaries. Include or exclude regions when the full image is not representative, and filter obvious artifacts before generating final statistics.
Measure size and shape at object level
Available particle metrics include area, perimeter, equivalent diameter, circularity, aspect ratio, solidity, extent, major/minor axes, orientation, Feret dimensions, eccentricity and related shape descriptors. When a valid calibration is attached, supported size metrics can be reported in physical units; otherwise the measurements remain in pixels.
Inspect distributions, not only one average
Review summary statistics and size-distribution views together with the object table and quality information. A mean diameter can hide a mixed population, poor segmentation or a long tail, so the distribution and rejected objects are part of the scientific review.
A short workflow
- Load the image and attach a calibration if physical units are required.
- Choose OpenCV thresholding or another available segmentation path and tune it against visible object boundaries.
- Set the ROI and review the mask, especially merged, border-touching or obviously spurious objects.
- Accept the objects you want to analyze, inspect distributions and shape metrics, then export the required data or report.
Output you can keep
Export particle-level data as CSV and generate supported PDF or HTML reports. Keep the source image, calibration and segmentation settings with the result so a reviewer can understand how the object measurements were produced.
Segmentation limits to check
- Otsu and other threshold methods work best when object and background intensity are sufficiently separable.
- Touching or overlapping particles are not guaranteed to be separated correctly; watershed can assist but merged boundaries still need visual review.
- Cellpose availability depends on the Local Engine and current deployment; do not base a required workflow on it without checking access first.
- Calibration, image resolution and segmentation quality directly affect physical size statistics.
Continue your work