Related Experiment Video
Updated: Sep 12, 2025
08:32
Using Computer-based Image Analysis to Improve Quantification of Lung Metastasis in the 4T1 Breast Cancer Model
Published on: October 2, 2020
6.4K
Comprehensive Benchmark Dataset for Pathological Lymph Node Metastasis in Breast Cancer Sections.
Xitong Ling1, Yuanyuan Lei2, Jiawen Li1
1Shenzhen International Graduate School, Tsinghua University, Shenzhen, 518071, China.
Scientific Data
|August 7, 2025
Summary
This study enhances the Camelyon dataset for computational pathology by improving whole slide image (WSI) quality and labels. This refined dataset advances AI development for more accurate cancer diagnostics.
Area of Science:
- Computational pathology
- Digital pathology
- Artificial intelligence in healthcare
Background:
- Whole slide images (WSIs) from optical microscopy scanning are crucial for computational pathology (CPath).
- AI-driven diagnostic support for WSI analysis is rapidly advancing.
- The Camelyon dataset is widely used for benchmarking but requires quality and label evaluation.
Purpose of the Study:
- To reprocess and enhance the Camelyon-16 and Camelyon-17 datasets for improved AI benchmarking.
- To address limitations in label quality, accessibility, and clinical relevance.
- To upgrade the cancer screening task and re-evaluate AI methods.
Main Methods:
- Reprocessed 1,399 WSIs from Camelyon-16 and Camelyon-17.
- Removed low-quality slides and corrected erroneous labels.
- Provided expert pixel annotations for tumor regions and upgraded the task to four classes (negative, micro-metastasis, macro-metastasis, ITC).
Main Results:
- A cleaned and enhanced Camelyon dataset with improved annotations.
- A re-evaluation of pre-trained pathology feature extractors and multiple instance learning (MIL) methods.
- Established a benchmark for advancing AI development in histopathology.
Conclusions:
- The enhanced Camelyon dataset provides a more robust benchmark for computational pathology.
- Improved annotations and task definition facilitate more accurate AI model development.
- This work contributes to advancing AI-driven diagnostic support in digital pathology.
