Evaluation of 3D Ultrasound Reconstruction and 2D/3D Segmentation for Neonatal Hip Dysplasia Screening

Wiebke Heyer, Katharina Ott, Christian Weihsbach, Reza Sorbi, Lisa Lange, Jürgen Lichtenstein, Anna Hell, Sebastian Lippross, Lasse Hansen, Mattias P Heinrich

Keywords: Hip Dysplasia, Ultrasound, Validation, Segmentation, 3D Reconstruction

TL;DR: Validation study of 3D reconstructed freehand ultrasound for neonatal hip dysplasia screening, comparing 2D and 3D deep-learning–based segmentation methods.

Abstract: Early detection of developmental dysplasia of the hip relies heavily on the correct acquisition and interpretation of ultrasound images. Yet, conventional single-plane imaging provides only a limited view of the neonatal hip, is operator-dependent and sensitive to probe orientation. In this study, we present a clinically oriented validation of a dual-sweep 3D ultrasound approach aimed at improving anatomical coverage and simplifying the diagnostic process. Our dataset comprises 50 optically tracked acquisitions and 150 untracked freehand sweeps from newborns, enabling the reconstruction of volumetric representations of the hip from standard handheld 2D ultrasound. We evaluate 2D and 3D nnU-Net–based segmentation models to quantify how volumetric context influences the delineation of key joint structures. Results demonstrate that the combination of 2D slice-based and 3D volumetric segmentation yields the most robust performance, particularly in cases with anatomical variability or suboptimal sweep direction. The study also highlights remaining challenges, including motion artefacts and inconsistent sweep trajectories, that affect reconstruction quality.

Primary Subject Area: Segmentation

Secondary Subject Area: Image Acquisition and Reconstruction

Registration Requirement: Yes

Reproducibility: https://github.com/MDL-UzL/JOINT

Visa & Travel: No

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Replace NNN with your OpenReview submission ID., Includes \documentclass{midl}, \jmlryear{2026}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package., Did not use the times package., Use the correct spelling and format, avoid Unicode characters, and use LaTeX equivalents instead., Any math in the title and abstract must be enclosed within $...$., Did not override the bibliography style defined in midl.cls and did not use \begin{thebibliography} directly to insert references., Avoid using \scalebox; use \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., No separate supplementary PDF uploads., Acknowledgements, references, and appendix must start after the main content.

Submission Number: 30