Loading...
 
Skip to main content
Scientific Volume Imaging B.V.

Scientific Volume Imaging

Deconvolution - Visualization - Analysis



Huygens' AI segmentation

Embedded within Huygens Object Analyzer, Huygens’ AI-driven segmentation leverages Meta’s Segment Anything Model 2 (SAM2)1 for high-performance automated object segmentation, including support for manual refinement. It handles both 2D and 3D datasets via an intuitive interface and can be combined with more classical segmentation approaches such as threshold and watershed. The segmented results can be directly used within Huygens' Object Analyzer for downstream quantitative analysis and visualization. Detailed object-based analysis in the Object Analyzer is also quickly performed after uploading segmentation results from tools like Cellpose2 and StarDist3. Curious about Huygens' other AI solutions? See here!


No programming required

Easily adjustable segmentation

Automatic and AI-guided manual segmentation

Built-in 3D visualization

Direct analysis with the Object Analyzer



AI-Based Segmentation of Overlapping Nuclei: Huygens automatic AI segmentation accurately separates overlapping objects by recognizing shape and context. This image of nuclei was deconvolved before AI segmentation, and kindly provided by Dr. Nicolas Fête, Laboratory of Stem Cells Dynamics, Swiss Federal Institute of Technology of Lausanne, Switzerland.

What can you use it for?


Automatic AI segmentation

The fully automatic segmentation is particularly effective for images where objects like cells or nuclei are densely clustered and do have well-defined boundaries. In such cases, the SAM1 algorithm can accurately distinguish individual structures with minimal errors. If the segmentation contains inconsistencies like slightly irregular borders or small missed areas, they can be easily corrected with minimal manual effort. This combination of efficient automation and easy manual refinement makes the AI segmentation in Huygens accurate and time-saving.

Nuclei Segmentation Using Automated AI: Automated segmentation enables efficient, high-throughput segmentation of nuclei with minimal user input, while still allowing the flexibility for manual refinement, significantly accelerating analysis workflows. The image was deconvolved before AI segmentation, and kindly provided by Dr. Nicolas Fête, Laboratory of Stem Cells Dynamics, Swiss Federal Institute of Technology of Lausanne, Switzerland.
Automatic Segmentation UI Update


Purkinje Seg


AI-guided manual segmentation

For more complex images, such as those containing numerous structures or background elements that are not relevant to the analysis, Huygens' AI-guided manual segmentation provides a more targeted approach. Instead of relying on traditional manual methods that require time-consuming "painting" or pixel-by-pixel adjustments, users can interact with the image in a much simpler way. By either clicking directly on the object of interest or drawing a bounding box around it, the SAM21 algorithm is able to recognize and segment the object automatically.

AI-Guided Manual Segmentation Segments the Cell Bodies of Purkinje Cells:

Huygens' AI-guided manual segmentation has the ability to distinguish detailed structures such as the cell bodies of Purkinje cells, even when surrounded by other tissue components. One simple click on the object is enough for the AI to segment it. AI-guided manual segmentation on highly pixelated raw image of brain tissue with Purkinje cells. Courtesy of Dr. Zsolt Csaba, Platform DHIM. Inserm, Robert Debre Hospital, Paris, France


3D segmentation

Huygens also offers AI-powered 3D segmentation, designed to accurately identify and separate complex structures within 3D images. Huygens' intuitive and user-friendly interface makes the entire process easy to navigate, with clear controls and visual feedback that guides users at every step. Combined with tools for quick correction and detailed visualization, Huygens offers a powerful, accessible solution for researchers working with 3D datasets.

AI segmentation of nuclei in a 3D cell cluster 3D view of nuclei in a cell cluster, the image was deconvolved before segmentation with Huygens' AI tool. Courtesy: Dr. Nicolas Fête, Laboratory of Stem Cells Dynamics, Swiss Federal Institute of Technology of Lausanne, Switzerland.

3Dseg90frames

How does it work?

Step 0: deconvolve and restore your data

AI segmentation performs significantly better on properly Nyquist sampled and deconvolved images. In this case a comparison was made between raw (left) and deconvolved (right) images of plant embryo's.

Raw data

Deconvolved data

Courtesy: Dr. Saiko Yoshida and Dr. Ton Timmers (Central Microscopy), Max Planck Institute for Plant Breeding Research, MPI, Cologne, Germany. Zeiss Airyscan image of a embryo of Arabidopsis thaliana, expressing nuclear and plasma membrane markers for observing 3D cellular morphology.

Step 1: Thresholds

Huygens offers two intuitive and easily adjustable thresholds that can be set to match the segmentation with the specific requirements of your experiment. By offering direct visual feedback Huygens helps ensure that the thresholds are set correctly, improving both the accuracy and biological relevance of the results.

Threshold on mimimum size

Size Threshold


A minimum object size threshold can be set to filter out irrelevant objects, ensuring the software only recognizes structures of meaningful size.

Threshold on intensity

Intensity Threshold


Use the slider to adjust the minimum intensity threshold, helping prevent background signal from being misidentified as valid objects. In this view, red highlights areas recognized as real signal, not background.
Images courtesy of Dr. Nicolas Fête, Laboratory of Stem Cells Dynamics, Swiss Federal Institute of Technology of Lausanne, Switzerland.

Step 2: Choose your segmentation method, fully automatic or AI-guided manual segmentation

Fully automatic segmentation works well for images with densely clustered but well-defined structures, accurately separating objects with minimal errors. Minor imperfections can be quickly corrected with minimal manual input, making the process efficient and precise.

For more complex images, Huygens' AI-guided manual segmentation offers a targeted alternative. Users can simply click or draw a box around the object, and the AI handles the rest—no tedious pixel-by-pixel editing required.

Automatic segmentation

AI-guided manual segmentation

Courtesy: Dr. Ioannis Alexopoulos ILH/CIGL Multiscale Imaging Platform, Justus Liebig University Giessen, Germany (left). Dr. Zsolt Csaba, Platform DHIM. Inserm, Robert Debre Hospital, Paris, France (right).

Step 3: Adjust the segmentation

Minor segmentation errors can be easily corrected, objects can be deleted, merged, made bigger, or smaller. The UI allows you to easily visualize and adjust the segmentation both in 2D and in 3D.

Adjustingsegmentation3D
Interactive Correction of AI-Based Nuclei Segmentation: Segmentation adjustments in nuclei. Segmentation mistakes were artificially introduced. Adjustments: fixing a segmentation by dragging the binding box, deleting an object, adding an object by setting a point prompt, merging two objects that should have been identified as one, merging two objects that are one in 3D. AI segmentation on deconvolved image of nuclei. Courtesy: Dr. Nicolas Fête, Laboratory of Stem Cells Dynamics, Swiss Federal Institute of Technology of Lausanne, Switzerland

Step 4: Analyse the data in the Object Analyzer


Once the segmentation process is complete, the resulting data can directly be analyzed within the Object Analyzer. This allows immediate access to powerful statistical tools for in-depth analysis of the segmented objects. Users can quickly assess key metrics such as object count, size distribution, shape characteristics, and distances between objects. In addition to the statistical outputs, AI-generated labels associated with each object are also preserved, providing valuable visualisation. Both the statistical data and the AI annotations can be effortlessly exported. Learn more about the Object Analyzer here.

Measurement of Nuclei Distance Post-Segmentation Using Object Analyzer: After segmentation, the distance between two nuclei can be easily measured within the Object Analyzer. AI segmentation on deconvolved image of nuclei. Courtesy: Dr. Nicolas Fête, Laboratory of Stem Cells Dynamics, Swiss Federal Institute of Technology of Lausanne, Switzerland
Analysis In OA

References

1.. Ravi N., Gabeur V., Hu Y., Hu R., Ryali C., Ma T., Khedr H., Rädle R., Rolland C., Gustafson L., Mintun E., Pan J., Vasudev Alwala K., Carion N., Wu C., Girshick R., Dollár P., Feichtenhofer C. SAM 2: Segment Anything in Images and Videos. https://ai.meta.com/research/publications/sam-2-segment-anything-in-images-and-videos/
2. Stringer C, Wang T, Michaelos M, Pachitariu M. Cellpose: a generalist algorithm for cellular segmentation. Nat Methods. 2021 Jan;18(1):100-106. doi: 10.1038/s41592-020-01018-x. Epub 2020 Dec 14. PMID: 33318659.
3. Schmidt, U., Weigert, M., Broaddus, C., Myers, G. (2018). Cell Detection with Star-Convex Polygons. In: Frangi, A., Schnabel, J., Davatzikos, C., Alberola-López, C., Fichtinger, G. (eds) Medical Image Computing and Computer Assisted Intervention – MICCAI 2018. MICCAI 2018. Lecture Notes in Computer Science(), vol 11071. Springer, Cham. https://doi.org/10.1007/978-3-030-00934-2_30