Scholarlink

<p>A novel deep learning technique has been developed to predict essential three-dimensional soft tissue landmarks of the facial structure in dentistry. Data integrity and accuracy are compromised when deep learning relies on converting 3D models into 2D representations. This study presents a three-dimensional soft tissue model of the face inspired by the image scaling index method (ISIM). It can accurately determine locations for area delineation and feature evaluation. The initial phase in an object recognition network typically involves the assessment of the extent of each organ. Three-dimensional models of different organs facilitate the anticipation of network learning for landmarks. The study results are consistent with previous methods for acquiring geometric knowledge. The recommended method reduces the mean error in local testing to 1.8. The mean error of the entire test dataset is also within this range; approximately 72% of the data is within 2.5. Furthermore, by delineating 32 landmarks, our approach outperforms all previous machine learning-based techniques. The results demonstrate the efficacy of 3D models for prediction, as the proposed method accurately predicts specific 3D facial soft tissue landmarks.</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

Leveraging a Novel Pre-Trained Neural Networks for Geometric Feature Analysis in 3D Tissue Images

  • 期刊论文

  • 发表日期：

    2025年9月12日

查看全文
  • Chandra Sekhar Koppireddy,
  • G. Siva Nageswara Rao

摘要

A novel deep learning technique has been developed to predict essential three-dimensional soft tissue landmarks of the facial structure in dentistry. Data integrity and accuracy are compromised when deep learning relies on converting 3D models into 2D representations. This study presents a three-dimensional soft tissue model of the face inspired by the image scaling index method (ISIM). It can accurately determine locations for area delineation and feature evaluation. The initial phase in an object recognition network typically involves the assessment of the extent of each organ. Three-dimensional models of different organs facilitate the anticipation of network learning for landmarks. The study results are consistent with previous methods for acquiring geometric knowledge. The recommended method reduces the mean error in local testing to 1.8. The mean error of the entire test dataset is also within this range; approximately 72% of the data is within 2.5. Furthermore, by delineating 32 landmarks, our approach outperforms all previous machine learning-based techniques. The results demonstrate the efficacy of 3D models for prediction, as the proposed method accurately predicts specific 3D facial soft tissue landmarks.