VACVPlaque: mobile photography of Vaccinia virus plaque assay with segmentation masks

<strong>How to Cite Us</strong>

De, T., Thangamani, S., Urbański, A., &amp; Yakimovich, A. (2025). A digital photography dataset for Vaccinia Virus plaque quantification using Deep Learning. <em>Scientific Data</em>, <em>12</em>(1), 719.

<pre>@article{de2025digital,
title={A digital photography dataset for Vaccinia Virus plaque quantification using Deep Learning},
author={De, Trina and Thangamani, Subasini and Urba{\'n}ski, Adrian and Yakimovich, Artur},
journal={Scientific Data},
volume={12},
number={1},
pages={719},
year={2025},
publisher={Nature Publishing Group UK London}
}</pre>

<strong>Data Description</strong>

The VACVPlaque dataset comprises spatially correlated objects, specifically virological plaques, which are circular phenotypes indicative of vaccinia virus (VACV) spread, and the wells of the assay plate. The virus plaque assay is a common method performed by infecting a monolayer of host cells (indicator cells) that are grown in the wells of assay plates or dishes. The host cells are infected with varying concentrations of a highly diluted virus inoculum. After an incubation period, typically around 48 hours, the cells are fixed with formaldehyde and stained with a dye to reveal the plaques or areas of cell death. By counting these plaques, researchers can calculate the number of infectious particles present in the original inoculum as described in [1].

This dataset consists of mobile photographs of 6-well tissue culture plates where the VACV plaque assay was conducted. The photographs were taken using two different mobile phones, resulting in 211, 8-bit RGB images with a resolution of 2448 x 3264 pixels. Each plate was photographed from two different perspectives using two different devices, meaning there are two images of the same plate but from different angles and devices.

To aid in the training of machine learning models, the dataset is divided into training, validation, and test subsets in a 70:20:10 ratio. To prevent data leaks, only one perspective of each image is included in the validation and test subsets. The training subset, which includes images from both perspectives, consists of 148 images.

<strong>File Description:</strong>

VACVPlaque_train.zip -&gt; train holdout

VACVPlaque_validation.zip -&gt; validation holdout

VACVPlaque_test.zip -&gt; test holdout

Each zip file contains:

images -&gt; {filename}.tif

plaque_masks -&gt; {filename}.tif

well_masks -&gt; {filename}.tif

<strong>References:</strong>

1. Dulbecco, Renato. "Production of plaques in monolayer tissue cultures by single particles of an animal virus." Proceedings of the National Academy of Sciences 38, no. 8 (1952): 747-752.

**引用说明**
De, T.、Thangamani, S.、Urbański, A.与Yakimovich, A.（2025）发表于《科学数据（Scientific Data）》的《基于深度学习的痘苗病毒（Vaccinia Virus, VACV）噬斑定量数字化摄影数据集》，刊载于第12卷第1期，页码719。

<pre>@article{de2025digital,
title={基于深度学习的痘苗病毒噬斑定量数字化摄影数据集},
author={De, Trina and Thangamani, Subasini and Urbański, Adrian and Yakimovich, Artur},
journal={Scientific Data},
volume={12},
number={1},
pages={719},
year={2025},
publisher={Nature Publishing Group UK London}
}</pre>

**数据集概况**
本数据集命名为VACVPlaque，包含空间关联的靶标对象，具体为病毒噬斑——痘苗病毒（Vaccinia Virus, VACV）扩散特征的圆形表型——以及培养板孔。
病毒噬斑实验是一种经典实验方法，具体操作为将宿主细胞（指示细胞）接种于培养板或培养皿的孔中形成单层，随后以不同浓度的高度稀释病毒接种物进行感染。经过约48小时的孵育期后，使用甲醛固定细胞并通过染料染色，以显现噬斑或细胞死亡区域。研究人员可通过计数噬斑数量，计算原始接种物中的感染性病毒颗粒数，具体方法详见文献[1]。

本数据集包含完成痘苗病毒噬斑实验的6孔组织培养板的智能手机拍摄照片。拍摄使用两款不同的智能手机，最终得到211张8位RGB图像，分辨率为2448×3264像素。每块培养板分别从两个不同视角、使用两款不同设备进行拍摄，即同一块培养板会生成两张不同角度、不同设备拍摄的图像。

为便于机器学习模型训练，本数据集按照70:20:10的比例划分为训练集、验证集与测试集三个子集。为避免数据泄露，验证集与测试集中仅包含每块培养板的单视角图像；训练集包含两种视角的图像，共计148张。

**文件说明**
VACVPlaque_train.zip → 训练留存集
VACVPlaque_validation.zip → 验证留存集
VACVPlaque_test.zip → 测试留存集

每个压缩包均包含以下内容：
images 文件夹：存储{filename}.tif格式的原始图像
plaque_masks 文件夹：存储{filename}.tif格式的噬斑掩码（mask）图像
well_masks 文件夹：存储{filename}.tif格式的培养板孔掩码图像

**参考文献**
1. Dulbecco, Renato. 《动物病毒单颗粒在单层组织培养中形成噬斑的方法》，刊载于《美国国家科学院院刊（Proceedings of the National Academy of Sciences）》第38卷第8期（1952年），页码747-752。