A Multi-Rater Benchmark for Perineuronal Nets Detection and Counting in Fluorescence Microscopy Images

The experimental images in the dataset have been obtained from mouse brain tissue with the protocol summarized below: A mouse was transcardially perfused with PBS and then 4% paraformaldehyde (PFA, wt/vol, dissolved in 0.1 M phosphate buffer, pH 7.4). The brain was post-fixed overnight in PFA 4%, then transferred to 30% sucrose (w/vol), 0.05% sodium azide (wt/vol) solution, and finally stored at 4°C. 50-µm coronal slices were cut on a freezing microtome (Leica), and free-floating sections were processed for immunohistochemistry with the following protocol. Incubation 2h at RT in blocking solution (3% bovine serum albumin (BSA) in PSB); incubation O/N at 4°C with a PBS solution of Fluorescein Wisteria floribunda lectin (Biotinylated Wisteria floribunda lectin, Vector Laboratories (1:200); rinse in PBS (3x10 mins at RT); incubation for 2 h 30 min with a PBS solution of fluorescent streptavidin (Streptavidin Alexa Fluor 488 conjugate, Thermo Fisher Scientific, 1:400). Slices were then rinsed in PBS (3x10 mins) coverslipped in mounting medium (VECTASHIELD antifade mounting medium, Vector Laboratories, cat.no. H-100), and stored at 4°C until acquisition. Images were acquired using an ApoTome.2 microscope (Zeiss, Oberkochen, Germany) and a 10X objective. Full images for each slice resulted from the stitching of multiple tiles that were automatically acquired. A single Z-plane (manually corrected for non planar deformations) was used to image each section. Image acquisition parameters were set in order to avoid as much saturation as possible in the brightest slices and were kept constant for the acquisition of all the experimental images. A single (PNN-SR) or seven different (PNN-MR) raters manually annotated the (x-y) positions of all the PNN they could visually detect in the slices.