PySpatial: A High-Speed Whole Slide Image Pathomics Toolkit.

Area of Science:

• Digital pathology
• Computational pathology
• Bioinformatics

Background:

• Whole Slide Image (WSI) analysis is vital for digital pathology, but traditional methods require complex, multi-step feature extraction pipelines.
• Existing pipelines, like those using CellProfiler, involve segmenting WSIs into patches, extracting features, and remapping them, leading to lengthy processing times.

Purpose of the Study:

• To introduce PySpatial, a novel, high-speed pathomics toolkit engineered for efficient WSI-level analysis.
• To streamline the conventional WSI feature extraction workflow by enabling direct analysis of computational regions of interest.

Main Methods:

• PySpatial employs rtree-based spatial indexing and matrix-based computation for efficient processing of computational regions.
• The toolkit bypasses redundant steps in traditional pipelines by operating directly on regions of interest, reducing overall workflow complexity.

Main Results:

• PySpatial demonstrated significant performance improvements on both Perivascular Epithelioid Cell (PEC) and Kidney Precision Medicine Project (KPMP) datasets.
• Nearly a 10-fold speedup was observed for small, sparse objects (PEC dataset), and a 2-fold speedup for larger objects like glomeruli and arteries (KPMP dataset).

Conclusions:

• PySpatial offers a substantial advancement in WSI analysis efficiency and accuracy for digital pathology.
• The toolkit's performance enhancements facilitate large-scale pathomics studies and broader applications in computational pathology research.