Confocal and STED Live F-actin dataset

Paired confocal and STED images of F-actin nanostructures in living neurons using the far-red fluorogenic dye SiR-Actin. This dataset was used to train and test the TA-GAN model for confocal-to-STED super-resolution of axonal and dendritic F-actin in living neurons (Resolution Enhancement with a Task-Assisted GAN to Guide Optical Nanoscopy Image Analysis and Acquisition). All images : 20 nm/pixel. Folders:<br> - train : 753 pairs of confocal/STED images (varying sizes)<br> - valid : 47 pairs of confocal/STED images (varying sizes)<br> - test_initial &amp; test_final : 84 confocal/STED pairs acquired before (initial) and 84 acquired after (final) control sequences where 15 confocal images of the full FOV (500 x 500 pixels = 10μm x 10μm) were acquired at 1 frame/minute. In addition to the confocal image, a sub-region (100 x 100 pixels = 2μm x 2μm) was selected outside the central ROI (300 x 300 pixels = 6μm x 6μm) and acquired with the STED modality at every time step; the signal decrease due to photobleaching effects can therefore be more prononced in the border region outside the central ROI.<br> - test_series : contains 149 series of images acquired with TA-GAN assistance using the change-based Dice coefficient threshold or the variability-based threshold. The test_series folder contains folders with names corresponding to "[date of acquisition]_cs[coverslip number]_ROI[selected region number]". Each folder contains three subfolders : input, full_STED, and initial_final_confocal. <br> - Input : this folder contains three-channel images for each of the 15 frames in the series. The first channel is the full FOV confocal image (10μm x 10μm), the second channel is the STED sub-region (2μm x 2μm) with zero-padding to match the shape of the FOV, and the third channel is a decision map of 0s and 1s, with the 1s indicating the position in the FOV of the STED sub-region.<br> - full_STED : this folder contains all the STED FOVs (10μm x 10μm) acquired when triggered by the TA-GAN assistance. The number of full_STED FOVs varies from 0 to 15 per region, with a mean of 2.8 STED images per series. The full FOV STED images acquired before (initialSTED.tif) and after (finalSTED.tif) the series of 15 frames are also included.<br> - initial_final_confocal : The full FOV confocal images acquired before (initialConfocal.tif) and after (finalConfocal.tif) the series of 15 frames.本数据集包含采用远红色荧光染料SiR-Actin标记的活神经元内F-肌动蛋白纳米结构的配对共聚焦与受激发射损耗显微镜（Stimulated Emission Depletion microscopy, STED）图像。该数据集被用于训练和测试任务辅助生成对抗网络（TA-GAN）模型，以实现活神经元轴突与树突内F-肌动蛋白的共聚焦图像到STED超分辨率图像的转换，对应研究论文标题为《基于任务辅助生成对抗网络的分辨率增强：指导光学纳米成像图像分析与采集》。所有图像的像素采样率均为20 nm/像素。文件夹结构如下： - train：753对共聚焦/STED图像（尺寸不一） - valid：47对共聚焦/STED图像（尺寸不一） - test_initial & test_final：包含分别采集于对照序列开始前（初始组）与结束后（终末组）的84对共聚焦/STED图像；该对照序列以1帧/分钟的速率采集了完整视场（Field of View, FOV）内的15幅共聚焦图像，完整视场尺寸为500×500像素，对应10μm×10μm。此外，在中心感兴趣区域（Region of Interest, ROI，300×300像素，即6μm×6μm）外选取了一个子区域（100×100像素，即2μm×2μm），并在每个时间步使用STED模态进行采集；因此光漂白效应导致的信号衰减在中心ROI外的边缘区域会更为显著。 - test_series：包含149组在TA-GAN辅助下采集的图像序列，这些序列采用基于变化的戴斯系数（Dice coefficient）阈值或基于变异性的阈值进行筛选。test_series文件夹下包含多个命名格式为"[采集日期]_cs[盖玻片编号]_ROI[选定区域编号]"的子文件夹。每个子文件夹均包含三个子目录：input、full_STED以及initial_final_confocal。 - input：包含该序列15帧图像各自对应的三通道图像。第一通道为完整视场共聚焦图像（10μm×10μm）；第二通道为经过零填充以匹配完整视场尺寸的STED子区域图像（2μm×2μm）；第三通道为0-1决策图，其中1值标记了STED子区域在完整视场内的位置。 - full_STED：包含所有在TA-GAN辅助触发下采集的完整视场STED图像（10μm×10μm）。每个序列的完整视场STED图像数量介于0至15幅之间，平均每个序列包含2.8幅STED图像。此外还包含该序列开始前采集的完整视场STED图像（initialSTED.tif）与结束后采集的完整视场STED图像（finalSTED.tif）。 - initial_final_confocal：包含该序列开始前采集的完整视场共聚焦图像（initialConfocal.tif）与结束后采集的完整视场共聚焦图像（finalConfocal.tif）。