Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Aug 27;14(1):19810.
doi: 10.1038/s41598-024-68084-5.

Neural shape completion for personalized Maxillofacial surgery

Stefano Mazzocchetti  1 , Riccardo Spezialetti  2 , Mirko Bevini  3 , Giovanni Badiali #  3   4 , Giuseppe Lisanti #  2 , Samuele Salti #  2 , Luigi Di Stefano #  2
Affiliations

Neural shape completion for personalized Maxillofacial surgery

Stefano Mazzocchetti et al. Sci Rep. .

Abstract

In this paper, we investigate the effectiveness of shape completion neural networks as clinical aids in maxillofacial surgery planning. We present a pipeline to apply shape completion networks to automatically reconstruct complete eumorphic 3D meshes starting from a partial input mesh, easily obtained from CT data routinely acquired for surgery planning. Most of the existing works introduced solutions to aid the design of implants for cranioplasty, i.e. all the defects are located in the neurocranium. In this work, we focus on reconstructing defects localized on both neurocranium and splanchnocranium. To this end, we introduce a new dataset, specifically designed for this task, derived from publicly available CT scans and subjected to a comprehensive pre-processing procedure. All the scans in the dataset have been manually cleaned and aligned to a common reference system. In addition, we devised a pre-processing stage to automatically extract point clouds from the scans and enrich them with virtual defects. We experimentally compare several state-of-the-art point cloud completion networks and identify the two most promising models. Finally, expert surgeons evaluated the best-performing network on a clinical case. Our results show how casting the creation of personalized implants as a problem of shape completion is a promising approach for automatizing this complex task.

Keywords: 3D deep learning; Maxillofacial surgery; Personalized medicine; Shape completion; Surgery planning.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1
Figure 1
(a) Examples for different Quality Score, (b) train, validation and test splits obtained considering the same proportion of samples w.r.t the quality scores.
Figure 2
Figure 2
Due to the internal points, two main volumes can be distinguished. The innermost is made of points that can be removed to simplify the point clouds since the most important outcome of the reconstruction process is the shape of the external surface. (a) Frontal view, (b) Left Lateral view, (c) Parietal view, (d) Basilar view.
Figure 3
Figure 3
Internal points removal pipeline: a Snapshots of the 3D model from different angles, (b-top) Depth maps of the snapshots, (b-bottom) Point clouds obtained from depth information, c Final result.
Figure 4
Figure 4
Defect Injection: (a, b) in blue, points in the area where a random point is selected for point clouds with quality score 5 (a) and 4 (b); (c) defect creation: the random point used as the center of the cuboid with square base is shown in red, while the selected points that will be removed to create the defect are in green; (d) some defects.
Figure 5
Figure 5
(a) Alignment after ShapeNet- like normalization. (b) Alignment after using a fixed scale factor m for normalization.
Figure 6
Figure 6
Complete pipeline for the proposed skull reconstruction framework.
Figure 7
Figure 7
(a) Reconstruction obtained from SnowflakeNet, (b) reconstruction obtained from PointAttN.
Figure 8
Figure 8
Qualitative reconstruction results.
Figure 9
Figure 9
Examples of failed reconstructions.
Figure 10
Figure 10
Example of a clinical case reconstruction.
See this image and copyright information in PMC

References

    1. Du, R. et al. A systematic approach for making 3d-printed patient-specific implants for craniomaxillofacial reconstruction. Engineering6, 1291–1301 (2020).10.1016/j.eng.2020.02.019 - DOI
    1. Chen, X., Xu, L., Li, X. & Egger, J. Computer-aided implant design for the restoration of cranial defects. Sci. Rep.7, 4199 (2017). 10.1038/s41598-017-04454-6 - DOI - PMC - PubMed
    1. Fei, B. et al. Comprehensive review of deep learning-based 3d point cloud completion processing and analysis. IEEE Transactions on Intelligent Transportation Systems (2022).
    1. Chang, A. X. et al. ShapeNet: An Information-Rich 3D Model Repository. Tech. Rep., Stanford University—Princeton University — Toyota Technological Institute at Chicago (2015). arXiv:1512.03012 [cs.GR]
    1. Chilamkurthy, S. et al. Development and validation of deep learning algorithms for detection of critical findings in head CT scans. arXiv preprint arXiv:1803.05854 (2018). - PubMed