Related Experiment Video
Updated: May 21, 2026
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
MetAssist 2.0: A Generalizable Artificial Intelligence Framework for Lymph Node Metastasis Detection Across Multiple
Javier Garcia-Baroja1, Bastian Dislich2, Philipp Zens3
1Institute of Tissue Medicine and Pathology, University of Bern, Bern, Switzerland; Graduate School for Cellular and Biomedical Sciences, University of Bern, Bern, Switzerland; Department of Digital Medicine, University of Bern, Bern, Switzerland.
MetAssist 2.0, an AI system, automates lymph node metastasis detection in pathology slides. This tool shows high accuracy across diverse cancer types, potentially reducing pathologist workload and improving cancer staging.
Area of Science:
- Oncology
- Pathology
- Artificial Intelligence
Background:
- Lymph node metastasis assessment is crucial for cancer staging but is labor-intensive and variable.
- Detecting micro-metastases and isolated tumor cells significantly impacts treatment decisions.
Purpose of the Study:
- To develop and validate MetAssist 2.0, an AI system for automated metastasis detection in lymph nodes.
- To assess the generalizability and clinical-grade performance of the AI system across various cancer types.
Main Methods:
- A modular AI system combining a pathology foundation model and transformer-based segmentation was developed.
- The system was trained on colorectal and upper gastrointestinal cancers and validated on 8,144 slides from seven cancer types and 14 cohorts.
- Few-shot fine-tuning was used to adapt the system to unseen cancer types with minimal data.
Main Results:
- MetAssist 2.0 achieved at least 90% sensitivity in 13 cohorts and 91% specificity in 11 cohorts.
- The system demonstrated high performance on challenging subtypes like mucinous adenocarcinoma and tumor deposits.
- As a triage tool for colorectal cancer, it reduced pathologist workload by up to 72% with 98% sensitivity and identified missed metastases.
Conclusions:
- MetAssist 2.0 demonstrates broad generalizability and near clinical-grade performance for automated metastasis detection.
- The AI system has the potential to be integrated into pathology workflows, improving efficiency and accuracy.
- The technology shows promise for enhancing cancer staging and guiding treatment decisions.
