Publications
When can you trust a computer’s outline of a tumour, and how does a busy clinician check it quickly? Everything below is one of three attempts at an answer.
T2 Fast and Reliable Dosimetric Contour QA, PhD thesis, University of Bern, 2025.
C conference · J journal · W workshop · T thesis, numbered oldest to newest.
Robust Segmentation Models
Which design choices survive a scan from a different hospital, and which quietly break. Watch them break.
- J5 The impact of U-Net architecture choices and skip connections on the robustness of segmentation across texture variations
paper · code · project page - J4 DreamOn: a data augmentation strategy to narrow the robustness gap between expert radiologists and deep learning classifiers
paper - C7 Do we really need that skip-connection? Understanding its interplay with task complexity
paper · code · talk · project page - W2 How do 3D image segmentation networks behave across the context versus foreground ratio trade-off?
paper · code · project page
Personalized Contour Review and Correction
Two experts outline the same tumour differently and both can be right. Telling genuine errors apart from reasonable disagreement.
- J7 Predicting the impact of target volume contouring variations on the organ at risk dose: results of a qualitative survey
paper · project page - J6 Efficient review of automatic contouring of OARs in the brain: a dual-layer quality assurance approach combining geometric and dosimetric validation
paper - W3 AutoDoseRank: automated dosimetry-informed segmentation ranking for radiotherapy
paper · code · project page - C8 Comparing the performance of radiation oncologists versus a deep learning dose predictor to estimate dosimetric impact of segmentation variations
paper · code · talk · project page - C6 Dose guidance for radiotherapy-oriented deep learning segmentation
paper - C5 ASTRA: atomic surface transformations for radiotherapy quality assurance
paper · code · talk · project page
Fast and Sensitive Dose Prediction
Predicting the radiation dose an outline would produce in seconds rather than hours, without losing the sensitivity that makes it worth having.
- J2 Deep-learning-based dose predictor for glioblastoma: assessing the sensitivity and robustness for dose awareness in contouring
paper · project page - C4 How sensitive are deep learning based radiotherapy dose prediction models to variability in organs at risk segmentation?
paper · code · talk · project page - C3 Evaluating a deep learning based 3D dose prediction model for quality assurance of organ at risk contours
paper
Other Work
Outside the three axes above. The older entries are from a previous life in diffusion MRI.
- J3 Orchestrating explainable artificial intelligence for multimodal and longitudinal data in medical imaging
paper - J1 PyRaDiSe: a Python package for DICOM-RT-based auto-segmentation pipeline construction and DICOM-RT data conversion
paper · code - C2 Optimal acquisition protocol for white matter fiber orientation mapping using generalized CSA-ODF reconstruction
paper - W1 Generalized constant solid angle ODF and optimal acquisition protocol for fiber orientation mapping
paper - T1 A generalized CSA-ODF model for fiber orientation mapping
record - C1 A novel device to monitor mobilization of fingers during treatment for stiffness of tendons
paper