Detecting and Segmenting 3D Objects within an RI Map

The 3D Cell Explorer from Nanolive is ideal when 3D images of living cells are to be created with very high spatial and temporal resolution at x,y: 180 nm; z: 400 nm; t: 1.7 sec. The capability of object detection and segmentation within these powerful images is very important in facilitating the type of cutting-edge research in biology that is based on imaging.

This capability is very complex, and experimenters need specific knowledge and competencies to deal with it. This article deals with the performance of one simple task, namely, the proper segmentation of mammalian cell nuclei in a 3D RI map with the use of FIJI software.


The 3D Cell Explorer is suited to the acquisition of a 3D map of the RI values within the objects under study (Cotte et al., 2013). The elegant design enables it to distinguish RI differences within 0.001, and puts it far beyond the next best on the market with respect to its sensitivity. This feature allows the experimenter to acquire an excellent view of various elements that make up the sample, such as different cell compartments.

This means that expertise is no longer key to identifying a mammalian cell nucleus in an RI map. In the current period. However, at the age of quantitative cell biology (Langen et al., 2015; Li et al., 2007) where seeing, feeling and manual image analysis is no more the norm, it is crucial that the computer crunches your image to extract meaningful object features organized in space, time and object identity (Frechin et al., 2015).

This article deals with key tools used for computation and basic image analytical knowledge. The example used here is the 3D segmentation of HeLa cell nuclei (Figure 1). The FIJI software (Schindelin et al., 2012; Schmid, Schindelin, Cardona, Longair, & Heisenberg, 2010; Schneider, Rasband, & Eliceiri, 2012) is required in addition to the STEVE software used for acquisition.

The process of exporting a good digital staining (D-stain) acquired with STEVE and improved with FIJI is also explained, as this leads to a high-quality 3D segmentation of the nuclei. Once this is understood, any object under study for which a D-stain exists can be redone the same way. The specific parameters should be adjusted for each experimental problem rather than taking the ones displayed in the current experiment as a hard and fast rule.

FIJI allows to do efficient 3D object segmentation

Figure 1. FIJI allows to do efficient 3D object segmentation


The tools needed for this experiment include the most current version of STEVE, the 3D Cell  Explorer software, and the latest version of FIJI (Schindelin et al., 2012; Schneider et al., 2012) ( accompanied by the 3D ImageJ suite shown in Figure 2. This suite is obtained by updating FIJI, as follows: click Help>Update… as seen in Figure 2, steps 1 and 2, then click manage update as in Figure 2, step 3.

Sometimes more than one update round may be necessary, and even one or more restarts as requested by FIJI, to reach the point where the process can be completed by simply checking the 3D ImageJ suite option seen in Figure 2, step 4. Now the window can be closed as shown in Figure 2, step 5. FIJI is restarted and the presence of the 3D and 3D viewer are both verified in the Plugin menu shown in Figure 2, step 6.

Installation of the 3D ImageJ suite

Figure 2. Installation of the 3D ImageJ suite

Lastly, the 3D Cell Explorer-generated acquisition file is required. A good-quality segmentation depends on good-quality acquisition in the first place. In this article four HeLa cells grown in a 3D gel matrix are used, which are at an early stage of cell division.

Exporting the D-Stain

Once a proper dataset acquisition is in place, the 3D refractive index (RI) map must be exported with the D-stain in the .tiff format. STEVE’s export tool is useful in exporting the dataset with an anyname.tiff stack. Following acquisition, load the dataset into STEVE if this has not been done already, using File>Load. Then go to File>Export as in Figure 3, steps 1 and 2.

This opens a panel, in which first RI volume is chosen as shown in Figure 3, step 3, and the format converted to Tiff as shown in Figure 3, step 4. This is exported after choosing a location and a file name, as shown in Figure 3, steps 5 and 6.

The process is reiterated as Figure 3, steps 7-12 show, to export the digital stain. These exports result in 96 slices in the Z dimension acquired. All of these slices are not always useful and unnecessary slices may just create noise in the 3D image. Thus they must be removed from the anyname.tiff file which can then be opened in a 3D visualization tool.

Export of D-stain and refractive index map from STEVE to .tiff

Figure 3. Export of D-stain and refractive index map from STEVE to .tiff

Preparation of Data

Following the reduction of the D-stain stack, Ds_anyname.tiff, to the data-containing slices, an 8-bit grayscale image must be produced from this. With FIJI, the procedure is as follows: go to Image>Type>8-bit to create a new stack (shown in Figure 4, steps 1-2). To bring down the noise, the data must be filtered using a smooth filter, which is 3D in this case to suit the data. To do this, go to Plugins>3D>3D Edge and Symmetry Filters as shown in Figure 4, steps 3-5, and put in the values in the 3D filter panels.

For Nanolive D-stains, the values shown in Figure 4 step 6 ought to be a good fit, but for small objects, the filtering may be avoided or small filter sizes may be used. The stack Ds_anyname.tiff can now be used for object detection.

Loading and preparation of the D-stain .tiff stack for further object detection

Figure 4. Loading and preparation of the D-stain .tiff stack for further object detection

Using 3D Plugin of FIJI to Detect Cell Nuclei

The signal contained in the Ds_anyname.tiff must be examined for the connected components which make up the objects, and this means analysis of voxel connectivity. Visual examination shows four nuclei clearly. It is also clear that some voxel blocks are not nuclei, though the D-stain process results in labeling them as nuclei parts.

These therefore need to be filtered out. Go to Plugins>3D>3D Simple Segmentation as in Figure 5, steps 1-3, to open a window (shown in Figure 5, step 4). Here three parameters need to be put in, but the lowest option must not be checked.

These include the following: a voxel is considered to be part of an object if the signal is over low threshold. Min size and Max size are parameters that delimit the least and greatest volumes in voxels that an object must have, if it is to be retained during the process of segmentation. These must be verified to be the right values before clicking OK since these determine segmentation quality. The four nuclei retained as objects are now virtually ready to be visualized.

Objects filtering and 3D segmentation of Nuclei

Figure 5. Objects filtering and 3D segmentation of Nuclei

The STEVE-aided D-stain has some imperfections which need to be rectified, as they could lead to the appearance of holes within the shapes of the objects detected, as shown in Figure 5. To fill these, click Plugins>3D>3D Binary Close Labels as shown in Figure 6, steps 1-3. This opens a window, as seen in Figure 6, step 4.

Two parameters must now be entered, RadiusXY and RadiusZ. The distance up to which a gap can be filled is defined. This varies from case to case, and the values shown here are not a definite rule. Press OK to create a new stack which describes the segmentation of nuclei in X, Y and Z all ready to be visualized.

Closing unwanted holes in 3D objects

Figure 6. Closing unwanted holes in 3D objects

It is possible to merge the 3D segmentation of the nuclei with the RI volume of the sample being studied so that the quality of segmentation may be understood more clearly. This helps to refine the parameters so that finally the process may be repeated with more finesse.

To do this, the RI stack in FIJI which has already been optimized by the elimination of z-slices that lack the right signal is loaded. Now the volume of the 3D nuclei and the RI volume are both open in FIJI as shown in Figure 7. Now go to Image>Color>Merge channels… as in Figure 7, steps 1-3.

This opens a third window shown in Figure 7, step 4. Now the volume to be attributed to each color must be selected. The segmentation is linked to the red channel as shown by the red star. The RI volume is linked to the gray channel as shown by the gray star. This is so that the RI image will appear in grayscale. Now “Create composite” is checked, and then click OK. The result will appear.

Merging the Nuclei segmentation and refractive index map in a single visualization

Figure 7. Merging the Nuclei segmentation and refractive index map in a single visualization


Cotte, Y., Toy, F., Jourdain, P., Pavillon, N., Boss, D., Magistretti, P., ... Depeursinge, C. (2013). Marker-free phase nanoscopy. Nature Photonics, 7(2), 113–117.

Frechin, M., Stoeger, T., Daetwyler, S., Gehin, C., Battich, N., Damm, E.-M., ... Pelkmans, L. (2015). Cell-intrinsic adaptation of lipid composition to local crowding drives social behaviour. Nature, 523(7558), 88–91.

Langen, M., Agi, E., Altschuler, D. J., Wu, L. F., Altschuler, S. J., & Hiesinger, P. R. (2015). The Developmental Rules of Neural Superposition in Drosophila. Cell, 162(1), 120–133.

Li, G., Liu, T., Tarokh, A., Nie, J., Guo, L., Mara, A., ... Wong, S. T. (2007). 3D cell nuclei segmentation based on gradient flow tracking. BMC Cell Biology, 8(1), 40.

Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., ... Cardona, A. (2012). Fiji: an open-source platform for biological-image analysis. Nature Methods, 9(7), 676–682.

Schmid, B., Schindelin, J., Cardona, A., Longair, M., & Heisenberg, M. (2010). A high-level 3D visualization API for Java and ImageJ. BMC Bioinformatics, 11(1), 274.

Schneider, C. A., Rasband, W. S., & Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image analysis. Nature Methods, 9(7), 671–675.

About Nanolive SA

Nanolive SA are scientists, working for scientists.

Their belief is that each and every Biologist, Researcher and Physician should be able to explore and interact instantly with living cells without damaging them.

Nanolive want to support the study of how living cells and bacteria work, evolve and react, thus building a solid base for new drugs and therapies, in order to enable breakthrough researches.

This is the reason why they have developed the 3D Cell Explorer.

Sponsored Content Policy: publishes articles and related content that may be derived from sources where we have existing commercial relationships, provided such content adds value to the core editorial ethos of News-Medical.Net which is to educate and inform site visitors interested in medical research, science, medical devices and treatments.

Last updated: May 6, 2019 at 10:32 AM


Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Nanolive SA. (2019, May 06). Detecting and Segmenting 3D Objects within an RI Map. News-Medical. Retrieved on October 18, 2019 from

  • MLA

    Nanolive SA. "Detecting and Segmenting 3D Objects within an RI Map". News-Medical. 18 October 2019. <>.

  • Chicago

    Nanolive SA. "Detecting and Segmenting 3D Objects within an RI Map". News-Medical. (accessed October 18, 2019).

  • Harvard

    Nanolive SA. 2019. Detecting and Segmenting 3D Objects within an RI Map. News-Medical, viewed 18 October 2019,

Other White Papers by this Supplier