GTTR Cell Fluorescence Quantification

Automatic cell and nucleus boundaries with the same per-cell measurements as the manual ImageJ workflow (total, nucleus and cytoplasm area; integrated densities; mean cytoplasm fluorescence; background). Upload the Zeiss ZEN RGB export (.tif, .jpg, .png) or the original .lsm file.

Image (ZEN export .tif / .jpg / .png, or .lsm)

Calibration

From the image metadata when available (LSM), otherwise the value below Use the value below

Pixels per micron

9.597 for this project (40x objective, zoom 1.5, 1024 px). The manual scale-bar measurements used 9.2 to 9.8.

Background

Estimate automatically (darkest empty 32x32 patch outside all cells) Enter a value

ZEN-style image labels

All cells: area and intensity mean on every cell and nucleus (full ZEN text for the listed cells) Cell numbers only (full ZEN text for the listed cells)

Cell numbers to label with the full ZEN text (comma-separated, from a previous run of the same image)

Predicted outlines (yellow = cell, cyan = nucleus; contrast-stretched for display)

Summary

ZEN-style annotation: white outlines, Area (um2) and Intensity Mean Value per cell and nucleus (rendered at 2x; download for full size)

Model: Cellpose-SAM fine-tuned for this project: cell model 'final_cells_v4_s050_e60' applied at 0.5x resolution, nucleus model 'final_nuclei_v4b_s050_e40' on the DAPI channel at 0.5x resolution, red-channel nucleus model 'final_nuclei_red_v4b_s050_e40' for files without DAPI (weights revision d50d40b). Fine-tuned on 68 images: 1,082 analyst-outlined cells recovered from the ImageJ overlays plus zero-shot Cellpose-SAM stand-ins for cells the analyst did not measure; nucleus models trained on the DAPI channel and on the red channel (for exports without DAPI). Outlines are moved 1.5 px (cells) and 2.5 px (nuclei) inwards to remove the systematic size bias relative to the analysts' outlines, fitted on images not used for training.

Measurements: areas are in square microns (pixel count divided by the square of the calibration). Integrated density is the sum of 8-bit red intensities over the region divided by the square of the calibration, as in ImageJ. MCF is the cytoplasm integrated density divided by the cytoplasm area, i.e. the mean red intensity of the cytoplasm.

Validated examples: click an image to run it with the calibration and background the analyst used; the analyst annotation then appears next to the prediction with the per-cell comparison.

Validated examples (held-out dishes; analyst annotation and Excel values included)