Built on Science.
Designed for the Clinic.
ContourAId is a web-based platform for radiotherapy contouring review, quality assessment, and dose analysis — developed at the intersection of clinical radiation oncology and medical artificial intelligence.
Reducing variability, improving outcomes
In radiotherapy, the quality of a treatment plan begins with the quality of the contour. How an oncologist delineates a tumour or an organ at risk directly shapes the dose delivered — and therefore the likelihood of local control and the risk of toxicity.
ContourAId exists to make contouring workflows more structured, more transparent, and more amenable to systematic quality review. By bringing versioning, comparison, and dose-awareness into a single environment, the platform supports researchers, clinical educators, and treatment teams in raising the standard of contour quality across the board.
A complete environment for radiotherapy contouring workflows
ContourAId integrates the tools that radiotherapy teams need for rigorous contour management — from initial upload through comparison, dose analysis, and export — in a single, access-controlled web application.
Multi-format Import & Export
Supports DICOM and NIfTI — the two dominant formats in radiotherapy research. Import existing contours and images, export to DICOM-RT Structure Set for treatment planning system compatibility.
Multiplanar Image Viewing
High-fidelity CT and MRI viewing in axial, sagittal, and coronal planes with dose overlay — giving clinicians the full anatomical context for every contouring decision.
Contour Versioning & History
Every contouring iteration is preserved as a named version with author and timestamp. Compare any two historical states, trace who made each change, and maintain a full audit trail of contouring decisions.
Pairwise Contour Comparison
Compare any two contour versions side-by-side — from the same observer at different times or from two independent delineators — with colour-coded overlap displays and Dice, Jaccard, and surface distance metrics.
Dose-Volume Analysis
Compute dose-volume histograms and standard dosimetric endpoints per structure. Understand how contouring decisions translate into dose coverage and organ sparing — with complete DVH curves and clinical endpoints on demand.
Contour Quality Metrics
Quantify agreement between any two contour versions using established geometric metrics — Dice similarity coefficient, Jaccard index, Hausdorff distance, and mean surface distance — providing an objective basis for contour review.
Grounded in peer-reviewed research
ContourAId is developed at the ARTORG Centre for Biomedical Engineering Research, University of Bern, in collaboration with the Department of Radiation Oncology, Inselspital Bern. The platform is motivated by and validated against the following body of work.
Deep-learning-based dose predictor for glioblastoma — assessing the sensitivity and robustness for dose awareness in contouring
Demonstrates that dose-prediction models can surface dosimetric consequences of contouring uncertainty in glioblastoma cases, motivating dose-aware contour review workflows.
A dual-layer quality assurance approach leveraging dose prediction for efficient review of automated contours
Introduces a two-tier QA framework combining geometric and dosimetric criteria to triage auto-segmented contours, reducing review burden while preserving safety.
How sensitive are deep learning-based radiotherapy dose prediction models to variability in organs at risk segmentation?
Quantifies how contouring uncertainty propagates into predicted dose distributions, providing evidence for the clinical relevance of segmentation quality.
Dose guidance for radiotherapy-oriented deep learning segmentation
Explores how dosimetric feedback can guide segmentation training, bridging the gap between geometric segmentation objectives and clinical treatment outcomes.
AutoDoseRank: Automated dosimetry-informed segmentation ranking for radiotherapy
Presents an automated ranking system for auto-segmented contours based on dosimetric impact, enabling prioritised clinical review of AI outputs.
ASTRA: Atomic surface transformations for radiotherapy quality assurance
Introduces geometric transformations as building blocks for systematic contour QA, enabling fine-grained characterisation of segmentation errors.
PyRaDiSe: A Python package for DICOM-RT-based auto-segmentation pipeline construction
Open-source tooling for building and evaluating radiotherapy segmentation pipelines using DICOM-RT — underpins the data infrastructure behind ContourAId.
Designed for radiotherapy professionals
Research Teams
Version, compare, and analyse expert contours with quantitative metrics. Benchmark contouring quality against reference delineations using a rigorous geometric and dosimetric framework.
Clinical Training Programmes
Trainees contour cases and receive immediate quantitative feedback. Supervisors review and compare against gold-standard references — making contour training measurable and structured.
Quality Review Teams
Compare contour versions with objective geometric metrics. Identify delineation discrepancies and support peer review with a reproducible, quantitative measurement framework.
Ready to explore ContourAId?
Request access to the platform for your research or training programme.
ContourAId is intended for research and educational use. It is not a certified medical device.