Scholarlink

<p>Digital dentistry increasingly relies on three-dimensional (3D) oral scans to improve crown design accuracy and enable patient-specific dental treatments. Nevertheless, existing systems are still limited with regard to the issues of automation, accuracy and proper handling of incomplete scan data. Therefore, this paper proposes ATCD-3DOS-DAPGAN, an automated framework that generates patient-specific tooth crowns from 3D oral scans using deep learning. Initially, input images are collected from intraoral 3D scanner dataset. Then the collected images undergo pre-processing using Chebyshev Rational Fractional-Order Filtering (CRFOF) to noise reduction and normalization. Then the pre-processed images are supplied to the segmentation stage using Context infused Swin-UNet (CIS-UNet) to segment the target teeth and adjacent teeth. The segmented image is fed into the Deep Adaptive Perceptual Generative Adversarial Network (DAPGAN) to generate a precise tooth crown design. The Starfish Optimization Algorithm (SOA) is used for optimizing the weight parameters of DAPGAN. The performance of the ATCD-3DOS-DAPGAN attains 28.16%, 27.23% and 28.12% higher accuracy, 27.09%, 26.19% and 29.14% higher precision and 23.33%, 20.19%, 21.10% lower Root Mean Square Error (RMSE)compared with existing techniques: convolutional neural network for automated tooth segmentation on intraoral scans (CNN-ATS-IS), clinically oriented automatic three-dimensional enamel segmentation using deep learning (COATES-DL) and development of automatic three dimensional model comparison for forensic identification and testing utilizing odontology data (A3D-FIT-OD) respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated and accurate tooth crown design using 3D oral scans with deep adaptive perceptual generative adversarial network

  • 期刊论文

  • 发表日期：

    2026年7月23日

查看全文
  • Lakshmi S.,
  • Shanmuganathan C.

摘要

Digital dentistry increasingly relies on three-dimensional (3D) oral scans to improve crown design accuracy and enable patient-specific dental treatments. Nevertheless, existing systems are still limited with regard to the issues of automation, accuracy and proper handling of incomplete scan data. Therefore, this paper proposes ATCD-3DOS-DAPGAN, an automated framework that generates patient-specific tooth crowns from 3D oral scans using deep learning. Initially, input images are collected from intraoral 3D scanner dataset. Then the collected images undergo pre-processing using Chebyshev Rational Fractional-Order Filtering (CRFOF) to noise reduction and normalization. Then the pre-processed images are supplied to the segmentation stage using Context infused Swin-UNet (CIS-UNet) to segment the target teeth and adjacent teeth. The segmented image is fed into the Deep Adaptive Perceptual Generative Adversarial Network (DAPGAN) to generate a precise tooth crown design. The Starfish Optimization Algorithm (SOA) is used for optimizing the weight parameters of DAPGAN. The performance of the ATCD-3DOS-DAPGAN attains 28.16%, 27.23% and 28.12% higher accuracy, 27.09%, 26.19% and 29.14% higher precision and 23.33%, 20.19%, 21.10% lower Root Mean Square Error (RMSE)compared with existing techniques: convolutional neural network for automated tooth segmentation on intraoral scans (CNN-ATS-IS), clinically oriented automatic three-dimensional enamel segmentation using deep learning (COATES-DL) and development of automatic three dimensional model comparison for forensic identification and testing utilizing odontology data (A3D-FIT-OD) respectively.