Related Experiment Video
Updated: Oct 3, 2025
10:59
Reconstruction of 3-Dimensional Histology Volume and its Application to Study Mouse Mammary Glands
Published on: July 26, 2014
14.6K
Integrating morphologic and molecular histopathological features through whole slide image registration and deep
Kevin Faust1,2, Michael K Lee3,4, Anglin Dent3
1Department of Computer Science, University of Toronto, 40 St. George Street, Toronto, ON M5S 2E4, Canada.
Neuro-Oncology Advances
|February 14, 2022
Summary
We developed an automated workflow using computer vision and deep learning to integrate histopathology data from multiple studies, aiding in diffuse glioma subclassification and uncovering new spatial patterns. This approach enhances neuro-oncology diagnostics.
Area of Science:
- Computational pathology
- Digital pathology
- Neuro-oncology
Background:
- Modern neuro-oncology relies on integrating morphologic and immunohistochemical patterns.
- Digital pathology and AI tools face challenges with large whole slide image (WSI) files and tissue section variability.
- Automating histomorphologic information integration across studies is difficult.
Purpose of the Study:
- To develop an automated workflow for aligning and integrating histopathological information from multiple studies.
- To demonstrate the utility of this workflow in molecular subclassification and discovery of spatial patterns in diffuse gliomas.
- To address challenges posed by WSI size and tissue preparation variability.
Main Methods:
- Coupling computer vision tools like scale-invariant feature transform (SIFT) with deep learning.
- Automated Whole Slide Image (WSI) alignment using SIFT.
- AI-driven analysis for molecular subclassification and spatial pattern discovery.
Main Results:
- SIFT workflow achieved automated WSI alignment in 91% of 107 cases (AUC=0.96).
- AI workflow differentiated brain tumor types and performed molecular subclassification of diffuse gliomas using biomarkers (IDH1-R132H, ATRX).
- Analysis revealed an unappreciated association between Olig2-positive and proliferative areas in gliomas (r=0.62).
Conclusions:
- The developed workflow is efficient and generalizable for advanced diagnostic and discovery tasks in surgical neuropathology.
- It automates immunohistochemically compatible analyses, inspired by neuropathologist approaches.
- This facilitates improved diagnostics and research in neuro-oncology.
