Nonrigid alignment of en bloc tissue specimen to resection bed to enhance correspondence for re-resection guidance

With approximately 890,000 new cases annually, head and neck squamous cell carcinoma has one of the highest recurrence rates among solid malignancies. Frozen section analysis is the standard of care for intraoperative margin assessment, yet accurately relocating detected positive margins on the resection bed remains challenging for surgeons. Significant mucosal shrinkage of resected specimens further complicates establishing correspondence between the specimen and the resection bed. To address this, we propose a nonrigid registration method to correct intraoperative deformation and reconstruct a 3D mesh of the specimen prior to shrinkage. Our approach involves acquiring structured light 3D scans of the resected specimen and RGBD camera acquired point clouds of the resection bed. The proposed registration method integrates regularized Kelvinlets displacement solutions with a linearized iterative boundary reconstruction approach. Using the collected data, the algorithm deforms the 3D mesh of the specimen to perform nonrigid alignment with the resection bed point cloud. After deformation correction, the specimen mesh can help surgeons find the correspondence between the resected specimen and the patient’s resection bed, thereby enhancing the accuracy of relocating positive margins. Evaluation in cadaver studies demonstrates improved specimen alignment with the resection bed, reducing errors from 3.5mm to 1.6mm (54%) and 9.7mm to 5.3mm (45%) according to Chamfer and Hausdorff distance metrics, respectively. We also observe sub-millimeter average fiducial registration error and 1.9mm average projected surface error. Our method can be used intraoperatively and integrated with other surgical guidance techniques to further enhance precision.