![[2602.12933] Deep-Learning Atlas Registration for Melanoma Brain Metastases: Preserving Pathology While Enabling Cohort-Level Analyses](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.12933] Deep-Learning Atlas Registration for Melanoma Brain Metastases: Preserving Pathology While Enabling Cohort-Level Analyses
Summary
This article presents a deep-learning framework for registering melanoma brain metastases (MBM) to a common atlas, enhancing cohort-level analyses while preserving pathological details.
Why It Matters
The study addresses the challenges of analyzing spatially heterogeneous melanoma brain metastases by providing a robust registration method that improves the accuracy of multi-centre research. This has implications for better understanding tumor behavior and treatment outcomes.
Key Takeaways
- Introduces a deep-learning framework for accurate registration of melanoma brain metastases.
- Preserves pathological details without the need for lesion masks, simplifying the analysis process.
- Demonstrates high registration accuracy across multiple datasets, facilitating reproducible research.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.12933 (cs) [Submitted on 13 Feb 2026] Title:Deep-Learning Atlas Registration for Melanoma Brain Metastases: Preserving Pathology While Enabling Cohort-Level Analyses Authors:Nanna E. Wielenberg, Ilinca Popp, Oliver Blanck, Lucas Zander, Jan C. Peeken, Stephanie E. Combs, Anca-Ligia Grosu, Dimos Baltas, Tobias Fechter View a PDF of the paper titled Deep-Learning Atlas Registration for Melanoma Brain Metastases: Preserving Pathology While Enabling Cohort-Level Analyses, by Nanna E. Wielenberg and 8 other authors View PDF HTML (experimental) Abstract:Melanoma brain metastases (MBM) are common and spatially heterogeneous lesions, complicating cohort-level analyses due to anatomical variability and differing MRI protocols. We propose a fully differentiable, deep-learning-based deformable registration framework that aligns individual pathological brains to a common atlas while preserving metastatic tissue without requiring lesion masks or preprocessing. Missing anatomical correspondences caused by metastases are handled through a forward-model similarity metric based on distance-transformed anatomical labels, combined with a volume-preserving regularization term to ensure deformation plausibility. Registration performance was evaluated using Dice coefficient (DSC), Hausdorff distance (HD), average symmetric surface distance (ASSD), and Jacobian-based measures. The method was applied to 209 MBM patients from th...
