Back to all papers

Freehand 3D ultrasound imaging of the femoral head and neck.

August 6, 2026pubmed logopapers

Authors

Michels AD,Wiersma OPC,Yu G,Thanamayooran A,Wong I,Adamson RBA

Affiliations (4)

  • School of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, Canada.
  • Division of Orthopaedic Surgery, Department of Surgery, Nova Scotia Health, Halifax, Canada.
  • Department of Orthopaedic Surgery, Faculty of Medicine, Dalhousie University, Halifax, Canada.
  • School of Biomedical Engineering, Faculty of Medicine, Dalhousie University, Halifax, Canada. [email protected].

Abstract

To evaluate the performance of a novel freehand 3D ultrasound system for assessing morphological deformities of the femoral head and neck associated with femoroacetabular impingement syndrome (FAIS). The system integrates an off-the-shelf optical tracker with a low-cost point-of-care ultrasound probe. Bone surfaces were segmented from B-mode images using a deep learning feature pyramid network trained on a publicly available dataset and refined via transfer learning with curated, manually segmented images. Deterministic algorithms were used to generate bone surface point clouds and meshes. System accuracy was assessed by comparing reconstructed bone surfaces to computed tomography (CT) in four cadaveric hips and two patient hips. The system's feasibility for clinical imaging was evaluated in five healthy volunteers. The ultrasound reconstructions achieved geometric agreement with CT, within-subject repeatability and within-subject reproducibility of 0.4-0.6 mm. Alpha angles derived from CT and ultrasound scan agreed to within 0.6°. Given that clinically significant bone deformities associated with FAIS are ≥ 2 mm and clinically significant differences in alpha angle are > 5°, the system's performance appears adequate for clinical morphometry. In vivo imaging demonstrated feasibility for clinical use. The combination of 3D optical tracking, ultrasound imaging, and deep learning-based segmentation using models trained on public datasets and enhanced by transfer learning can produce bone surface maps of the femoral head with sufficient accuracy to resolve clinically significant femoral head deformities. The system is suitable for clinical deployment.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.