Fluorescent Neuronal Cells Dataset

Category: Themed Directories
Tags: datasets, neuroscience, computer-vision
Provider: Alma Mater Studiorum – Università di Bologna
Source: Dataset page and download

An image dataset of fluorescent neuronal cells intended as a testbed for developing and benchmarking machine learning and image analysis methods in neuroscience.

Overview

The Fluorescent Neuronal Cells dataset is a collection of high-resolution fluorescence microscopy images of mice brain slices. It is designed as a benchmark and testbed for computer vision and deep learning methods applied to biomedical imaging, especially for tasks involving neuronal cell detection and counting.

A newer version of this dataset is available: alternative version link.

Features

• Domain & Purpose

  • Focus on biological imaging rather than natural images.
  • Intended to benchmark and develop computer vision and deep learning methods in neuroscience and biomedical imaging.
  • Aims to support:
    • Research in biomedical-related fields, where popular pre-trained models often perform poorly.
    • Methodological research in deep learning, addressing the specific requirements of fluorescence microscopy images.

• Data Contents

  • 283 images of fluorescent neuronal cells.
  • Images are high-resolution: 1600 × 1200 pixels each.
  • Images depict mice brain slices acquired by fluorescence microscopy.
  • Neurons are highlighted with a yellow marker: Cholera Toxin sub-unit b (CTb).

• Annotations / Ground Truth

  • Pixel-wise segmentation labels for stained neurons.
  • Ground-truth masks are black-and-white images:
    • Background: pixel value 0 (black).
    • Neurons: pixel value 255 (white).
  • Labels generated via a hybrid approach combining:
    • Semi-automatic semantic segmentation.
    • Manual semantic segmentation.

• Potential Applications

  • Semantic segmentation of neurons.
  • Object detection of neuronal cells.
  • Object counting / cell counting.
  • General benchmarking for deep learning in microscopy and neuroscience.

• File & Access Details

  • Distributed as a compressed archive: fluocells.zip.
  • Approximate size: 414 MB.
  • Direct download: https://amsacta.unibo.it/id/eprint/6706/1/fluocells.zip.

• Licensing & Usage

  • Licensed under Creative Commons Attribution – ShareAlike 4.0 (CC BY-SA 4.0).
    License details: http://creativecommons.org/licenses/by-sa/4.0/

• Related Publication

  • Morelli, R. et al., 2021. Automating cell counting in fluorescent microscopy through deep learning with c-ResUnet. Scientific Reports.
    DOI: https://doi.org/10.1038/s41598-021-01929-5

• Funding Acknowledgments

  • University of Bologna (RFO 2018).
  • European Space Agency (Research agreement collaboration 4000123556).

Authors / Contributors

Affiliation for all listed authors: University of Bologna.

• Luca Clissa – ORCID: 0000-0002-4876-5200
• Roberto Morelli – ORCID: 0000-0001-5090-9026
• Fabio Squarcio – ORCID: 0000-0002-6033-1042
• Timna Hitrec – ORCID: 0000-0002-9296-3482
• Marco Luppi – ORCID: 0000-0002-9462-5014
• Lorenzo Rinaldi – ORCID: 0000-0001-9608-9940
• Matteo Cerri – ORCID: 0000-0003-3556-305X
• Roberto Amici – ORCID: 0000-0002-9692-2215
• Stefano Bastianini – ORCID: 0000-0003-2468-1704
• Chiara Berteotti – ORCID: 0000-0002-4143-9445
• Viviana Lo Martire – ORCID: 0000-0001-8696-0835
• Davide Martelli – ORCID: 0000-0001-6523-9598

(Author list is truncated in the provided content; see the source page for the complete list if needed.)

Pricing

• Cost: Free to download and use under the CC BY-SA 4.0 license.

Additional Notes

• This dataset is included in curated lists of machine learning datasets, emphasizing its usefulness as a benchmark for deep learning in biomedical imaging.
• Users are encouraged to cite the associated Scientific Reports paper when using the dataset in academic work.