Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Frequency and Prognostic Significance of Genetic Abnormalities in a Subgroup of Patients With Intermediate-Risk Neuroblastoma: A SIOPEN Study.

JCO precision oncology·2026
Same author

Two-Stage vs Single-Stage Ilizarov Application in High-Risk Patients: Impact on Early Complications and Resource Utilization.

Foot & ankle international·2026
Same author

Development of a photocatalytic membrane screening reactor (PMSR) for standardized evaluation of immobilized photocatalytic support materials.

MethodsX·2026
Same author

Challenges in Diagnosing High-grade B-cell Lymphoma, NOS: Poor Interobserver Agreement on Its Morphologic Definition-An LLMPP Study.

The American journal of surgical pathology·2026
Same author

New biological insights into osteosarcoma-lessons from single cell sequencing studies.

Cancer metastasis reviews·2026
Same author

A targeted circulating tumor DNA landscape of copy number aberrations in large B-cell lymphomas.

Leukemia·2026

Related Experiment Video

Updated: Nov 11, 2025

Deep Learning-Based Segmentation of Cryo-Electron Tomograms
10:25

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

9.9K

Evaluation of Deep Learning Architectures for Complex Immunofluorescence Nuclear Image Segmentation.

Florian Kromp, Lukas Fischer, Eva Bozsaky

    IEEE Transactions on Medical Imaging
    |March 30, 2021
    PubMed
    Summary

    Instance-aware segmentation for nuclear images is challenging. Deep learning models like Cellpose and Mask R-CNN show superior performance on complex images, especially when trained with artificially generated data.

    More Related Videos

    Deep Neural Networks for Image-Based Dietary Assessment
    13:19

    Deep Neural Networks for Image-Based Dietary Assessment

    Published on: March 13, 2021

    9.6K
    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
    04:48

    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

    Published on: November 30, 2022

    3.1K

    Related Experiment Videos

    Last Updated: Nov 11, 2025

    Deep Learning-Based Segmentation of Cryo-Electron Tomograms
    10:25

    Deep Learning-Based Segmentation of Cryo-Electron Tomograms

    Published on: November 11, 2022

    9.9K
    Deep Neural Networks for Image-Based Dietary Assessment
    13:19

    Deep Neural Networks for Image-Based Dietary Assessment

    Published on: March 13, 2021

    9.6K
    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
    04:48

    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

    Published on: November 30, 2022

    3.1K

    Area of Science:

    • Computational biology
    • Medical image analysis
    • Artificial intelligence in microscopy

    Background:

    • Instance-aware segmentation of nuclear images is crucial for quantitative analysis.
    • Deep learning methods show promise but require systematic comparison on complex immunofluorescence images.
    • Annotated fluorescence nuclear image datasets are scarce, limiting deep learning model training.

    Purpose of the Study:

    • To systematically evaluate and compare deep learning architectures and conventional algorithms for nuclear image segmentation.
    • To assess the effectiveness of a novel strategy for generating artificial images to augment training datasets.
    • To determine the sufficiency of undergraduate-annotated images for training segmentation models.

    Main Methods:

    • Evaluation of U-Net, U-Net ResNet, Cellpose, Mask R-CNN, and KG instance segmentation against conventional watershed and graph-based methods.
    • Development and application of a strategy for creating artificial fluorescence nuclear images.
    • Comparative analysis using F1 scores and Dice scores on complex fluorescence nuclear images.

    Main Results:

    • Instance-aware architectures (Cellpose, Mask R-CNN) outperformed U-Net and conventional methods in F1 scores.
    • U-Net architectures achieved higher mean Dice scores.
    • Artificial image generation improved recall and F1 scores, particularly for complex images.
    • Mask R-CNN and Cellpose achieved top F1 scores under specific training and testing conditions.
    • Undergraduate-annotated images proved sufficient for training effective instance-aware segmentation models.

    Conclusions:

    • Instance-aware deep learning models, particularly Cellpose and Mask R-CNN, are effective for complex nuclear image segmentation.
    • Artificial data augmentation can significantly enhance segmentation performance.
    • The quality of annotations from non-experts is adequate for training robust nuclear segmentation models.