Related Experiment Video
Updated: Jul 15, 2025
06:47
Expanding the Comprehension of the Tumor Microenvironment using Mass Spectrometry Imaging of Formalin-Fixed and Paraffin-Embedded Tissue Samples
Published on: June 29, 2022
2.2K
Accelerating Cancer Histopathology Workflows with Chemical Imaging and Machine Learning
Kianoush Falahkheirkhah1,2, Sudipta S Mukherjee1, Sounak Gupta3
1Beckman Institute for Advanced Science and Technology, University of Illinois Urbana-Champaign, Urbana, Illinois.
Cancer Research Communications
|September 29, 2023
Summary
This study introduces a new method combining stimulated Raman scattering (SRS) microscopy and artificial intelligence (AI) to rapidly create high-quality virtual pathology images from fresh-frozen prostate tissue, significantly speeding up cancer diagnosis.
Area of Science:
- Biomedical imaging
- Computational pathology
- Chemical imaging
Background:
- Traditional histopathology is resource-intensive, involving extensive tissue processing, sectioning, and staining.
- Emerging chemical imaging techniques like stimulated Raman scattering (SRS) microscopy offer direct molecular composition analysis, potentially streamlining workflows.
- Artificial intelligence (AI) can enhance image quality and analysis in pathology.
Purpose of the Study:
- To integrate SRS microscopy into a pathology workflow for rapid chemical information acquisition from minimally processed prostate tissue.
- To develop AI-driven computational methods for generating virtual H&E images from SRS data.
- To assess the diagnostic utility and impact on pathologist agreement of AI-generated virtual stained images compared to traditional methods.
Main Methods:
- Stimulated Raman scattering (SRS) microscopy was used to acquire chemical information from intact, thick fresh-frozen prostate tissues.
- Optical sectioning was employed to generate images from multiple planes within the tissue.
- A deep learning-based processing pipeline was utilized to create virtual hematoxylin and eosin (H&E) stained images.
- The method was extended to generate archival-quality images rapidly, comparable to formalin-fixed, paraffin-embedded (FFPE) processing.
Main Results:
- Virtual H&E images were generated from fresh-frozen prostate tissue in minutes, bypassing traditional processing steps.
- The quality of the virtual stained images was found to be diagnostically useful for pathologists.
- Interpathologist agreement on prostate cancer grading remained unaffected by the use of virtual stained images.
- The method's ability to preserve lipids and small molecules allowed for assessment of their utility in determining cancer grade.
Conclusions:
- The integration of SRS microscopy and AI significantly accelerates pathology workflows by reducing tissue processing time and complexity.
- AI-generated virtual stained images provide diagnostically valuable information comparable to traditional methods.
- This approach offers novel capabilities for rapid tissue assessment in pathology, with potential applications in cancer grading and molecular analysis.
