Healthcare
Earth Observation
Industry 5.0
Diagnostics
Medical Image Management
Breast Imaging
Enhancing Auto-Segmentation Research with 600+ Mammography Cases
December 23, 2025
The University Hospital of Basel
The University Hospital of Basel leads TotalSegmentator, an open-source research initiative dedicated to advancing clinically relevant medical image segmentation for the scientific community.
Breast cancer affects one in eight women worldwide.
Accurately segmenting breast carcinoma remains challenging due to variability in breast density, imaging conditions, and the need to reliably distinguish malignant tumours from benign findings.
Project Goals
- Enable reliable segmentation of breast carcinoma across diverse breast densities and imaging conditions.
- Support radiologists and researchers in differentiating malignant tumours from benign findings.
- Contribute clinically reliable annotated datasets to strengthen open-source breast imaging models.
Introduction
The University Hospital of Basel leads TotalSegmentator, an open-source research initiative dedicated to advancing clinically relevant medical image segmentation for the scientific community.
Main Challenge
Accurately segmenting breast carcinoma remains challenging due to variability in breast density, imaging conditions, and the need to reliably distinguish malignant tumours from benign findings.
Goals
- Enable reliable segmentation of breast carcinoma across diverse breast densities and imaging conditions.
- Support radiologists and researchers in differentiating malignant tumours from benign findings.
- Contribute clinically reliable annotated datasets to strengthen open-source breast imaging models.
Advancing open-source breast cancer segmentation research
Built on Ingedata’s radiologist-led workflows, the project delivered high-quality annotations to support more consistent and reliable breast cancer detection research.
Project Impact
Improved consistency in breast carcinoma segmentation
Accelerated development of auto-segmentation models
Stronger foundations for earlier and more reliable breast cancer detection
Clinically Precise Breast Carcinoma Annotation at Scale
Expert Breast Imaging Annotation
Deployed radiologists with dedicated expertise in breast imaging to ensure clinically accurate interpretation and annotation of mammography data.
0
1
Expert Breast Imaging Annotation
Density-Aware Image Analysis
Applied consistent annotation protocols across a wide range of breast densities and imaging conditions to minimise variability and improve model robustness.
0
2
Density-Aware Image Analysis
Malignancy Differentiation
Carefully distinguished malignant tumours from benign findings and normal anatomical variations to support clinically meaningful segmentation outcomes.
0
3
Malignancy Differentiation
Large-Scale Case Annotation
Manually segmented more than 600 breast carcinoma cases using the client’s annotation platform, ensuring consistency across the entire dataset.
0
4
Large-Scale Case Annotation
Specialist Clinical Review
Delivered over 45 hours of specialist radiologist review to validate annotations and ensure accuracy, consistency, and research-grade quality.
0
5
Specialist Clinical Review
IngeData's team of expert radiologists delivered precise, consistent annotations across more than 600 cases, with impressive efficiency. This project represents a major step forward in developing robust, clinically relevant AI models for breast cancer detection.
Jakob Wasserthal
Research Scientist
University Hospital Basel (TotalSegmentator)
