Related Experiment Video
Updated: Oct 13, 2025
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
382
Development of pathological reconstructed high-resolution images using artificial intelligence based on whole slide
Yang Deng1, Min Feng1,2,3, Yong Jiang3
1Laboratory of Pathology Key Laboratory of Transplant Engineering and Immunology NHC, West China Hospital Sichuan University Chengdu China.
Medcomm
|November 12, 2021
Summary
This study introduces a deep learning method to enhance 20x whole slide images (WSIs) to high-resolution 40x without losing details. This innovation addresses storage and transmission challenges in digital pathology, improving AI-assisted cancer diagnosis.
Area of Science:
- Digital Pathology
- Artificial Intelligence in Medicine
- Computational Pathology
Background:
- Digital pathology (DP) utilizes whole slide images (WSIs) for AI-assisted cancer diagnosis.
- Current WSI digitization often uses 20x or 40x objectives.
- 40x objective scans create large files, hindering DP adoption due to storage and transmission issues.
Purpose of the Study:
- To develop a deep learning-based method for reconstructing high-resolution (HR) 40x WSIs from 20x WSIs.
- To ensure the reconstructed HR WSIs retain both global and local pathological features.
- To validate the reconstructed HR WSI quality and consistency with actual 40x scans.
Main Methods:
- A novel deep learning algorithm was developed to upscale 20x WSIs to 40x resolution.
- The method was tested on WSI data from 100 uterine leiomyosarcomas and 100 adult granulosa cell tumors.
- Quantitative image quality metrics (PSNR, SSIM, BRISQUE) were used for evaluation.
Main Results:
- The reconstructed HR WSI achieved a peak signal-to-noise ratio (PSNR) of 42.03.
- Structural similarity (SSIM) was measured at an excellent 0.99.
- Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE) score was 49.22, indicating high quality.
- Consistency between reconstructed and actual HR images was confirmed.
Conclusions:
- The deep learning-based reconstructed HR imaging is a reliable technique for digital pathology slides.
- This method effectively converts 20x WSIs to 40x resolution without significant feature loss.
- The technique offers a scalable solution for clinical pathology, overcoming storage and transmission limitations of high-resolution WSIs.
