3D histology data of mouse prostates with manually annotated tumor hotspots

This dataset contains six sparse 3D histology stacks of registered, H&E-stained mouse prostate samples with manually annotated tumor hotspots. The data has been processed to suit SparStVR application. Each stack has approximately 30 sections with 50 micrometer distance between sections, and highest magnification is 10x. The dataset includes high resolution png images, subsampled tumor and sample masks and correspondingly sized H&E-stained images, and Cycle-GAN generated IHC and PAS stains plus experimental translations to forested ortho images. In addition, we provide feature maps for nucleus amount, computed from hematoxylin channel with two different tile sizes. Data is from mice heterozygous for tumor suppressor Pten+/-. Prostates were collected at the age of 11 months. Samples were fixed in PAXgene paraffin and cut into 5 micrometer thick slices. Every tenth section was H&E-stained and imaged with Zeiss Axioskop40 microscope (Carl Zeiss MicroImaging, NY, USA) using CCD color camera (QICAM Fast; QImaging, Canada) and motorized specimen stage (Märzhäuser Wetzlar GmbH, Germany). Surveyor imaging system (Objective Imaging, UK) controlled the automated image acquisition. Images were registered using Elastic Stack Alignment (ESA) plugin of Fiji (ImageJ).