Scalable Image Analysis for Next-Generation 3D Histopathology
Scholler J, Jonsson J, Jordá-Siquier T, Gantar I, Batti L, Cheeseman BL, Pagès S, Sbalzarini IF, Lamy CM. Efficient image analysis for large-scale next generation histopathology using pAPRica. bioRxiv [Preprint] 2023: 2023.01.27.525687. doi: 10.1101/2023.01.27.525687.
Background: Next-generation histopathology enables volumetric imaging of entire organs at high resolution, but the resulting datasets – often reaching terabyte or even petabyte scales – pose major challenges for data processing and analysis. Current methods require high-performance computing infrastructure and are too slow for clinical use or high-throughput research.
Hypothesis: This study tested the hypothesis that an Adaptive Particle Representation (APR)-based image analysis pipeline can efficiently process large-scale, multi-channel histological datasets acquired with light-sheet microscopy – specifically ClearScope – on standard workstations, while maintaining analytical accuracy.
Methods: The authors developed pAPRica, an APR-native processing framework that performs stitching, segmentation, registration and visualization directly on particle-based representations. A key test involved imaging a 1.7×2.7×0.3 cm3 human brain tissue block from an Alzheimer’s disease case using a ClearScope. The microscope employed dual digitally-scanned light sheets and chromatic aberration correction to capture high-resolution, multi-channel data (3.9 TB total). Each tile was converted to APR format in real time, allowing efficient segmentation and alignment during acquisition.
Results: APR conversion enabled a 117-fold reduction in data size and a 66-fold increase in processing speed relative to voxel-based methods. Segmentation of Aβ plaques achieved >99.9% object-matching accuracy compared to voxel-based ground truth. The framework also enabled cortical depth and density analysis of plaques. Performance was consistent across synthetic petascale datasets.
Conclusions: The integration of ClearScope imaging with APR-native analysis through pAPRica provides a scalable and accurate solution for real-time processing of large 3D histological datasets. This approach supports whole-organ analysis and has direct applicability in both clinical diagnostics and large-scale research projects.
