Neurite Measurement for the Whole Image

Introduction

This Fiji batch macro is designed to process neuron images with extensive clustering, especially those with low-contrast neurites. By utilizing the Local Thickness [1] and Skeletonize [2] plugins, we have developed a workflow for whole-image neurite measurement. The automatically saved Excel file provides the total neurite length and cell count for the entire image.

ver 2.1.0 – Updated on 2025/6/5 with the following changes: 

1. Set the binary background to white to prevent errors during skeletonization.
2. Added: Enhance Contrast, Gaussian Blur, and Otsu-based thresholding (user adjustable) on neuron mask.
3. Implemented dialog windows for selecting folders.
4. Image names are more descriptive and user-friendly.
5. Added 3 checkpoints to prevent runtime errors.

Examples

1. The confocal image of SH-SY5Y cells was acquired using high-content imaging. (courtesy of Dr. Ling-Wei Hsin (Department of Pharmacy, National Taiwan University).

Description 

1. This is a batch IJM script.
2. The demo image contains two channels: SH-SY5Y cells (green) and DAPI (blue).
3. The script begins by splitting the channels and renaming them accordingly.
4. Neurite Measurement
5. Then creating neuron mask by using the RenyiEntropy[3] thresholding method.
   • The neuron mask is duplicated, and local thickness is applied to approximate the soma mask.
   • The neuron mask is skeletonized, and the soma mask is subtracted to isolate the neurites.
   • The total length of the neurites is measured.
6. Cell Count
   • Otsu[4] thresholding is applied to the DAPI channel, and the result is converted to a mask.
   • The DAPI mask is multiplied with the normalized neuron mask to remove non-neuron cells.
   • The individual nucleus is identified by StarDist[5].
7. The total cell number is counted, and the average neurite area per cell is calculated.
8. All measurements are saved in a collection table.
9. A composite image is generated to visualize the results: raw neurons in white, segmented nucleus in the glasbey on dark channel, and neurites in red.
10. Both the composite image and measurement results are saved in the same output file.

Instruction

1. Place the image in the same directory for batch analysis. Also, create a null file to serve as the output file.
2. Drag the script and the demo image to Fiji.
3. Press “Run” and choose the input and output file respectively.
4. The collection table will be saved as an Excel file.

Tutorial

Acknowledgements

Thank to Dr. Shao-Chun, Peggy, Hsu, and Ms. Anchi Luo for their invaluable teaching and guidance!
Demo images are captured by Yu-Hsuan Lin and courtesy from Dr. Ling-Wei Hsin (Departrment of Pharmacy, National Taiwan University).

This work was supported by National Science and Technology Council NSTC 113-2320-B-002-076 to Shao-Chun Hsu.

Reference

1. R. P. Dougherty and K.-H. Kunzelmann, “Computing Local Thickness of 3D Structures with ImageJ,” in Microscopy & Microanalysis 2007 Meeting, Ft. Lauderdale, FL, USA, Aug. 2007.
2. T. Y. Zhang and C. Y. Suen, “A fast parallel algorithm for thinning digital patterns,” Communications of the ACM, vol. 27, no. 3, pp. 236–239, 1984.
3. P. Sahoo, C. Wilkins, and J. Yeager, “Threshold selection using Renyi’s entropy,” Pattern Recognition, vol. 30, no. 1, pp. 71–84, Jan. 1997, doi: 10.1016/S0031-3203(96)00065-9.
4. N. Otsu, “A threshold selection method from gray-level histograms,” IEEE Transactions on Systems, Man, and Cybernetics, vol. 9, no. 1, pp. 62–66, 1979, doi: 10.1109/TSMC.1979.4310076.
5. M. Weigert and U. Schmidt, “Nuclei Instance Segmentation and Classification in Histopathology Images with StarDist,” in 2022 IEEE International Symposium on Biomedical Imaging Challenges (ISBIC), Kolkata, India, 2022, pp. 1–4, doi: 10.1109/ISBIC56247.2022.9854534.