Three-dimensional (3D) surface reconstruction of the breast has become increasingly
relevant in clinical applications such as surgical planning, postoperative assessment, and aesthetic
evaluation. Recent advances in mobile device-based 3D scanning, particularly using tablet-based
applications, have enabled accessible and low-cost acquisition of patient-specific surface data. These
approaches offer an attractive alternative to traditional imaging modalities, which are often
expensive, time-consuming, or require controlled environments.
However, meshes obtained from mobile scanning devices are typically affected by noise, incomplete
geometry, and limited anatomical fidelity. Recent work [1] has demonstrated the feasibility of
generating breast surface meshes using iPad-based scanning applications tested on a female torso
phantom with silicone breasts. Despite the good results, the study also highlights limitations in
geometric accuracy and consistency of anatomical landmarks.
In parallel, advances in deep learning (DL) for 3D geometry processing, particularly mesh refinement
and reconstruction, have shown promising results in improving surface quality and structural
coherence. A recent line of work [2] proposes DL-based methods for refining coarse or noisy meshes
into anatomically plausible and geometrically consistent representations.
Despite these developments, it remains unclear whether such refinement techniques can
meaningfully improve clinically relevant measurements derived from mobile-acquired breast meshes,
such as inter-landmark distances or symmetry metrics.

To investigate whether iRBSM, the DL-based mesh refinement method proposed in [2], can
improve the anatomical accuracy of breast surface reconstructions obtained from mobile device
scans. More specifically, the study will:
• Reconstruct breast surface meshes from mobile scanning data (based on existing pipelines).
• Apply state-of-the-art mesh refinement models to improve surface quality.
• Evaluate the impact of refinement on clinically relevant anatomical distances and geometric
consistency.
• Compare refined and non-refined meshes using quantitative metrics.

Literature Review (ETOC: 2 months): Review existing methods for 3D breast surface
reconstruction using mobile devices. Study mesh processing and refinement techniques, with
a focus on deep learning-based approaches.
• Implementation (ETOC: 6 months): Reproduce or adapt the existing iRSBM model for breast
mesh reconstruction from mobile scans. Implement or integrate mesh refinement models
from recent literature, ensuring compatibility between reconstruction outputs and
refinement inputs. Compare original and refined meshes and perform statistical analysis to
assess improvements in anatomical fidelity. (Optionally) Explore validation on a limited set of
representative cases under appropriate constraints.


[1] Botti, E. Jansen, B. Ballen-Moreno, F. Kapila, A. Brahimetaj, R. Evaluating the Accuracy and
Repeatability of Mobile 3D Imaging Applications for Breast Phantom Reconstruction. Sensors 2025,
25, 4596.
[2] Weiherer, M., von Riedheim, A., Brébant, V., Egger, B., Palm, C. (2025). iRBSM: A Deep
Implicit3DBreastShape Model. In: Palm, C., et al. Bildverarbeitung für die Medizin 2025. BVM 2025.
Informatik aktuell. Springer Vieweg, Wiesbaden. https://doi.org/10.1007/978-3-658-47422-5_11

(Mandatory) qualifications:
• Following an MSc in a field related to one or more of the following: Computer
Science, Biomedical Engineering, Applied Computer Science - Digital Health.
• Strong programming skills (Python).
• Ability to write scientific reports and communicate research results at conferences in English.
• Prior experience with tools such as Blender is preferred.