GeDi and ICP Registration of Jawbone CT and Intraoral Scan Data with Pseudo-IOS Ground Truth

A recent paper on arXiv (2608.07564) presents a novel method for coarse-to-fine registration aimed at aligning CT scans of jawbones with data from intraoral scanners (IOS). This alignment is essential in digital dentistry and oral surgery, as it integrates the internal bone structure with detailed dental surface geometry. A significant challenge is that these two imaging modalities overlap only minimally, and their actual correspondence remains uncertain, hindering precise quantitative assessment. To address this, the authors propose a pseudoIOS evaluation framework, creating a point cloud from CT data that mimics an intraoral scan, thus establishing a known transformation. Their method utilizes a domain-generalizable local descriptor (GeDi) for initial global alignment and the iterative closest point (ICP) algorithm for precise registration, enhancing accuracy in dental surgical planning and prosthetic design. The paper can be accessed on arXiv with the identifier 2608.07564.