Fully automatic integration of dental CBCT images and full-arch intraoral impressions with stitching error correction via individual tooth segmentation and identification

TL;DR

This work tackles multimodal dental image fusion by automatically registering high-resolution intraoral scans (IOS) with CBCT volumes to recover both crown/gingival surfaces and tooth roots. It introduces a four-part pipeline: TSIM-IOS for per-tooth segmentation on IOS, TSIM-CBCT for per-tooth segmentation on CBCT, global-to-local tooth registration with a tooth-aware ICP refinement, and per-tooth stitching error correction of IOS using CBCT geometry. The method achieves clinically relevant accuracy, robust to metal artifacts, and provides a componentized jaw-tooth-gingiva model suitable for occlusal analysis and surgical guide production, thereby reducing MAR concerns and traditional impression workflow. Experimental results show per-tooth identification and segmentation performance in the mid-to-high 90s percentile, and final landmark/surface errors on the order of $\sim$ $2.20\times 10^2\,\mu\mathrm{m}$ and $4.72\times 10^2\,\mu\mathrm{m}$, with post-correction improvements to $\sim$ $1.12\times 10^2\,\mu\mathrm{m}$ and $3.02\times 10^2\,\mu\mathrm{m}$, highlighting practical utility for digital dentistry. The approach leverages $SE(3)$ transforms, Fast Point Feature Histograms, and tooth-specific ICP to minimize nonoverlapping regions and maximize accurate tooth-to-tooth correspondences, contributing a robust, automated solution for IOS–CBCT fusion.

Abstract

We present a fully automated method of integrating intraoral scan (IOS) and dental cone-beam computerized tomography (CBCT) images into one image by complementing each image's weaknesses. Dental CBCT alone may not be able to delineate precise details of the tooth surface due to limited image resolution and various CBCT artifacts, including metal-induced artifacts. IOS is very accurate for the scanning of narrow areas, but it produces cumulative stitching errors during full-arch scanning. The proposed method is intended not only to compensate the low-quality of CBCT-derived tooth surfaces with IOS, but also to correct the cumulative stitching errors of IOS across the entire dental arch. Moreover, the integration provide both gingival structure of IOS and tooth roots of CBCT in one image. The proposed fully automated method consists of four parts; (i) individual tooth segmentation and identification module for IOS data (TSIM-IOS); (ii) individual tooth segmentation and identification module for CBCT data (TSIM-CBCT); (iii) global-to-local tooth registration between IOS and CBCT; and (iv) stitching error correction of full-arch IOS. The experimental results show that the proposed method achieved landmark and surface distance errors of 112.4 $μ$m and 301.7 $μ$m, respectively.

Figures (8)

• Figure 1: Overall flow diagram of the proposed method consisting of four parts; tooth segmentation and identification from IOS and CBCT data, global-to-local tooth registration of IOS and CBCT, and stitching error correction in IOS. Therefore, the proposed method integrates IOS and CBCT images into one coordinate system while improving the accuracy of full-arch IOS.
• Figure 2: Results of TSIM-IOS and -CBCT, respectively. The indicated numbers represent mandibular teeth by the universal notation. (a) Individual IOS teeth and their split gingiva parts, and (b) CBCT teeth containing unexposed wisdom teeth.
• Figure 3: Qualitative comparison results of registration methods. (a) MR, (b) CPD, (c) FPFH, (d) FPFH followed by ICP, and (e) the proposed method. In the 3D visualization of the results, the colors on the IOS teeth represent distances between the IOS and CBCT tooth surfaces. On the 2D CT slice images, the red contours are cross sections of the aligned IOS models cut along the corresponding CT slice.
• Figure 4: Correspondence pairs of FPFH-based methods. The image on the left shows poor matching from the FPFH method without TSIM. On the other hand, the image on the right shows modest correspondences between the teeth obtained by TSIM.
• Figure 5: Qualitative results before and after correction of four selected evaluation data. The yellow and red lines represent contours of the IOS models with the proposed registration and correction methods, respectively. The two contours almost overlap, but the differences appear at the end of the arches.