Augmented Reality-based Guidance with Deformable Registration in Head and Neck Tumor Resection

TL;DR

The paper tackles inaccurate margin relocation in head and neck cancer by introducing a thickness-aware deformable registration that uses both the pre-resection external surface and the post-resection cavity, integrated with AR-guided visualization. It presents a Kelvinlet-based registration with strain-energy regularization and an isotropic scaling term, producing a deformed specimen mesh that aligns with the cavity and is overlaid onto the surgical field via an AR headset. Evaluations across skin, buccal, and tongue specimens show improved TRE over rigid and prior deformable methods, with tongue cases showing the strongest gains; a pilot AR-guided study reduces target relocation error from 9.8 cm to 4.8 cm and yields a ~46% improvement in accuracy. The work demonstrates a practical, 30-minute workflow on a standard laptop and HMD, indicating potential for meaningful clinical impact in reducing recurrence by improving intraoperative margin guidance, while acknowledging data limitations and the need for further validation.

Abstract

Head and neck squamous cell carcinoma (HNSCC) has one of the highest rates of recurrence cases among solid malignancies. Recurrence rates can be reduced by improving positive margins localization. Frozen section analysis (FSA) of resected specimens is the gold standard for intraoperative margin assessment. However, because of the complex 3D anatomy and the significant shrinkage of resected specimens, accurate margin relocation from specimen back onto the resection site based on FSA results remains challenging. We propose a novel deformable registration framework that uses both the pre-resection upper surface and the post-resection site of the specimen to incorporate thickness information into the registration process. The proposed method significantly improves target registration error (TRE), demonstrating enhanced adaptability to thicker specimens. In tongue specimens, the proposed framework improved TRE by up to 33% as compared to prior deformable registration. Notably, tongue specimens exhibit complex 3D anatomies and hold the highest clinical significance compared to other head and neck specimens from the buccal and skin. We analyzed distinct deformation behaviors in different specimens, highlighting the need for tailored deformation strategies. To further aid intraoperative visualization, we also integrated this framework with an augmented reality-based auto-alignment system. The combined system can accurately and automatically overlay the deformed 3D specimen mesh with positive margin annotation onto the resection site. With a pilot study of the AR guided framework involving two surgeons, the integrated system improved the surgeons' average target relocation error from 9.8 cm to 4.8 cm.

Figures (5)

• Figure 1: The standard of care and proposed tumor resection workflows. In the standard of care, the surgeon relies purely on verbal instructions to locate the positive margin, leading to imprecise re-resection. AR guidance with deformable registration provides additional visual guidance.
• Figure 2: a) A range of input data for the deformation registration task. b) The expected output of the deformation task.
• Figure 3: General Workflow of the AR system. a) From post-resection cavity point cloud extracted from the RGB-D camera, we obtain the pose of the post-resection cavity and the ArUco marker. b) The transformation between the ArUco marker and the post-resection cavity. c) The surgeon wearing the HMD can visualize the annotated specimen mesh, automatically overlaid on the post-resection cavity. d) The surgeon places a surgical pin on the target to evaluate end-to-end accuracy.
• Figure 4: The 3D specimen with rigid registration to post-resection cavity (first row), deformable registration with a post-resection cavity guidance (second row), and the proposed deformable registration with additional pre-resection external surface guidance (third row), respectively.
• Figure 5: The overall TRE (mm) of four different registration methods. The proposed method significantly outperforms the rigid and similarity registrations, which was not possible with the previous method. *: $p<0.05$.