Scholarlink

Statement of problem <p>Zirconia implants offer acceptable osseointegration in pre-clinical studies and may present clinical advantages in aesthetics, plaque affinity and biocompatibility; however, their fracture resistance remains debated.</p> Purpose <p>This study aims to evaluate the predictability of fractures in zirconia implants—specifically those resulting from structural changes due to prolonged functional loading—through the application of a machine learning (ML) algorithm developed using radiomic features.</p> Material and methods <p>Micro-computed tomography (micro-CT) images with a resolution of 10.1&#xa0;μm were acquired from 48 zirconia implants subjected to chewing simulation and 25 surviving implants. Following image post-processing, a comprehensive set of radiomic features was extracted using a radiomics platform designed to detect fracture-related changes. Nine machine learning algorithms were implemented and evaluated for their diagnostic performance.</p> Results <p>The area under the curve (AUC) values for the training datasets ranged from 0.950 to 1.000, while AUC values for the test datasets ranged from 0.336 to 0.722. Among the algorithms tested, the Random Forest (RF) model demonstrated the highest diagnostic accuracy for detecting zirconia implant fractures from micro-CT imaging.</p> Conclusions <p>In an ex-vivo chewing-simulation model, a Random-Forest classifier (AUC = 0.72) modestly predicted fracture-prone zirconia implants from micro-CT radiomics. After prospective CBCT validation, such models might assist but not replace clinical decision-making.</p> Clinical implications <p>The radiomic features and machine learning algorithm developed in this study, particularly when adapted to cone beam computed tomography (CBCT) imaging, may serve as a valuable tool in clinical practice for predicting implant outcomes and guiding decision-making regarding long-term implant maintenance.</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

Predicting fracture risk in Y-TZP zirconia implants using radiomics-based artificial intelligence

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月3日

查看全文
  • Fehmi Gönüldaş,
  • Arzu Alan,
  • Arda Buyuksungur,
  • Bora Akat,
  • Pelin Atalay Seçkiner,
  • Ata Bugra Keşan,
  • Kaan Orhan

摘要

Statement of problem

Zirconia implants offer acceptable osseointegration in pre-clinical studies and may present clinical advantages in aesthetics, plaque affinity and biocompatibility; however, their fracture resistance remains debated.

Purpose

This study aims to evaluate the predictability of fractures in zirconia implants—specifically those resulting from structural changes due to prolonged functional loading—through the application of a machine learning (ML) algorithm developed using radiomic features.

Material and methods

Micro-computed tomography (micro-CT) images with a resolution of 10.1 μm were acquired from 48 zirconia implants subjected to chewing simulation and 25 surviving implants. Following image post-processing, a comprehensive set of radiomic features was extracted using a radiomics platform designed to detect fracture-related changes. Nine machine learning algorithms were implemented and evaluated for their diagnostic performance.

Results

The area under the curve (AUC) values for the training datasets ranged from 0.950 to 1.000, while AUC values for the test datasets ranged from 0.336 to 0.722. Among the algorithms tested, the Random Forest (RF) model demonstrated the highest diagnostic accuracy for detecting zirconia implant fractures from micro-CT imaging.

Conclusions

In an ex-vivo chewing-simulation model, a Random-Forest classifier (AUC = 0.72) modestly predicted fracture-prone zirconia implants from micro-CT radiomics. After prospective CBCT validation, such models might assist but not replace clinical decision-making.

Clinical implications

The radiomic features and machine learning algorithm developed in this study, particularly when adapted to cone beam computed tomography (CBCT) imaging, may serve as a valuable tool in clinical practice for predicting implant outcomes and guiding decision-making regarding long-term implant maintenance.