Data retrieval from archival renal biopsies using nonlinear microscopy

Cahill, Lucas1; Yoshitake, Tadayuki1; Rosen, Milan2; Weber, Timothy1 ; Fujimoto, James1; Rosen, Seymour2

Published Feb 14, 2024 on Dryad. https://doi.org/10.5061/dryad.pzgmsbctz

Data files

Feb 14, 2024 version files 227.99 GB

Abstract

Thorough examination of renal biopsies may improve understanding of renal disease. Imaging of renal biopsies with fluorescence nonlinear microscopy (NLM) and optical clearing enables three-dimensional (3D) visualization of pathology without microtome sectioning. Archival renal paraffin blocks from 12 patients were deparaffinized and stained with Hoechst and Eosin for fluorescent nuclear and cytoplasmic/stromal contrast, then optically cleared using benzyl alcohol benzyl benzoate (BABB). NLM images of entire biopsy fragments (thickness range 88-660 µm) were acquired using NLM with fluorescent signals mapped to an H&E color scale. Cysts, glomeruli, exudative lesions, and Kimmelstiel-Wilson nodules were segmented in 3D and their volumes, diameters, and percent composition could be obtained. The glomerular count on 3D NLM volumes was high indicating that archival blocks could be a vast tissue resource to enable larger-scale retrospective studies. Rapid optical clearing and NLM imaging enables more thorough biopsy examination and is a promising technique for analysis of archival paraffin blocks.

README: Image datasets accompanying "Data retrieval from archival renal biopsies using nonlinear microscopy"

https://doi.org/10.5061/dryad.pzgmsbctz

The submission contains all source fluorescence microscopy datasets used to produce figures for the manuscript entitled, "Data retrieval from archival renal biopsies using nonlinear microscopy" PLOS ONE (Cahill et al., 2024). The source files are streams of 2D tiles, which when stitched together, form 3D microscopic representations of archival renal biopsies. For convenience, we have also included pre-stitched, full-resolution & bit depth, 2-channel stacks.

Dataset name: Manuscript figure number: Description
LC-6-2: Figure 1: Image of a biopsy from a patient with advanced diabetic nephropathy.
LC-7-2-rdo2: Figures 3 & 4: Image of a biopsy showing focal and segmental glomerulosclerosis with features of collapsing glomerulopathy
LC-9-1-rdo: Figure 5: Image of a biopsy showing crescentic IgA nephropathy.
LC-10-1: Figure 2: Image of a biopsy from a patient diagnosed with diabetic nephropathy and IgA nephropathy.
LC-11-2-rdo: Figure 6: Image of a biopsy showing tubular necrosis.
LC-13-1: Figure 10: Image of a biopsy from a patient diagnosed with lithium nephropathy.

Description of the data and file structure

Each dataset contains a stream of tiles and a pre-stitched 3D stack. Both files are OME-TIFF hyperstacks.

Image tiles have dimensions 2×1024×1024 (Channels×Height×Width) and are encoded in 16 bit unsigned integers. The time axis/dimension is used to represent tile index. The tiles are acquired in a boustrophedon pattern to span the entire specimen area.

Stitched hyperstacks have dimensions Z×2×Y×X, where Z is the total number of slices (depths) acquired, Y is the total number of pixels in height, and X is total number of pixels width. Pixel spacing information is included in the image metadata.

The channel order is eosin followed by Hoechst.

Code/Software

Both hyperstacks can be opened with ImageJ/Fiji using "Import"->"Bio-Formats" and selecting "View stack with:" "Hyperstack". Depending on your system memory and the hyperstack file size, you may also need to check "Use virtual stack". In composite color mode, eosin should appear as the red channel and Hoechst should appear as the green channel. Alternatively, datasets can be opened in programmatic languages, for instance in Python with the tifffile package. Images can be recolored to emulate the appearance of hematoxylin-eosin stain under brightfield illumination using previously published algorithms.