Spatial Analysis for AI-segmented Histopathology Images: Methods and Implementation

2025-12-05

AI Summary

To address the lack of efficient, accessible tools for quantitative spatial analysis of cellular organization in AI-segmented histopathological images, this study introduces SASHIMI—the first comprehensive single-cell spatial analytics platform specifically designed for AI segmentation outputs. SASHIMI integrates 27 spatial statistical and topological features, including proximity metrics, grid-based similarity measures, spatial autocorrelation indices, and persistent homology descriptors, and enables reproducible, real-time analysis via a web-based interactive visualization interface. Validated on cohorts of oral potentially malignant disorders and non-small cell lung cancer, SASHIMI identified multiple spatial biomarkers significantly associated with overall survival, demonstrating its utility in prognostic modeling. This work bridges a critical methodological gap in AI-driven spatial histomics, advancing standardization and clinical translation of tumor microenvironment quantification.

Technology Category

• Computer Vision: Segmentation
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Web Mining and Content Analysis: Web data visualization

Abstract

Quantitatively characterizing the spatial organization of cells and their interaction is essential for understanding cancer progression and immune response. Recent advances in machine intelligence have enabled large-scale segmentation and classification of cell nuclei from digitized histopathology slides, generating massive point pattern and marked point pattern datasets. However, accessible tools for quantitative analysis of such complex cellular spatial organization remain limited. In this paper, we first review 27 traditional spatial summary statistics, areal indices, and topological features applicable to point pattern data. Then, we introduce SASHIMI (Spatial Analysis for Segmented Histopathology Images using Machine Intelligence), a browser-based tool for real-time spatial analysis of artificial intelligence (AI)-segmented histopathology images. SASHIMI computes a comprehensive suite of mathematically grounded descriptors, including spatial statistics, proximity-based measures, grid-level similarity indices, spatial autocorrelation measures, and topological descriptors, to quantify cellular abundance and cell-cell interaction. Applied to two cancer datasets, oral potentially malignant disorders (OPMD) and non-small-cell lung cancer (NSCLC), SASHIMI identified multiple spatial features significantly associated with patient survival outcomes. SASHIMI provides an accessible and reproducible platform for single-cell-level spatial profiling of tumor morphological architecture, offering a robust framework for quantitative exploration of tissue organization across cancer types.

Problem

• Develops a tool for spatial analysis of AI-segmented histopathology images
• Quantifies cellular organization and interactions to understand cancer progression
• Identifies spatial features linked to patient survival outcomes in cancers

Innovation

• Browser-based tool for real-time spatial analysis of AI-segmented histopathology images
• Computes comprehensive spatial descriptors including statistics and topological features
• Identifies spatial features associated with patient survival outcomes in cancer datasets