Scholarlink

<p>Dental caries detection from radiographs is a crucial task in dental diagnostics. With the advancement of artificial intelligence (AI), convolutional neural networks (CNNs) have gained prominence in automated image-based diagnosis. This review specifically investigates the use of CNN models for caries detection in bitewing radiographs, aiming to highlight the current state of the art, identify methodological weaknesses such as data leakage, and provide future research recommendations. A systematic literature review was conducted by screening major academic databases including ScienceDirect, SpringerLink, Wiley, and Google Scholar. A total of 30 studies published between 2020 and 2025 were analyzed. The selected articles were evaluated based on problem type (classification, segmentation, object detection), CNN architecture, dataset characteristics, validation strategies, performance metrics, and technical innovations. The findings reveal that the majority of studies focused on object detection and segmentation tasks, with limited use of external validation and publicly available datasets. Only a few works introduced technical novelty in architectural design. Data leakage due to improper dataset partitioning and augmentation practices was identified in several studies, raising concerns about model generalizability. Most models reported high performance using metrics such as accuracy, precision, and recall, but lacked justification for metric selection and reproducibility assurance. This review emphasizes the need for rigorous data management practices in dental AI research. Future studies should adopt patient-level data splitting, utilize external validation sets, and prioritize model interpretability and open science practices. The review serves as a critical resource for guiding the development of clinically robust CNN-based caries detection systems.</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

Convolutional Neural Network Applications in Dental Diagnostics: A Focused Review on Bitewing-Based Caries Detection

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • Abdurrahman Ali Öncel,
  • Enes Ayan

摘要

Dental caries detection from radiographs is a crucial task in dental diagnostics. With the advancement of artificial intelligence (AI), convolutional neural networks (CNNs) have gained prominence in automated image-based diagnosis. This review specifically investigates the use of CNN models for caries detection in bitewing radiographs, aiming to highlight the current state of the art, identify methodological weaknesses such as data leakage, and provide future research recommendations. A systematic literature review was conducted by screening major academic databases including ScienceDirect, SpringerLink, Wiley, and Google Scholar. A total of 30 studies published between 2020 and 2025 were analyzed. The selected articles were evaluated based on problem type (classification, segmentation, object detection), CNN architecture, dataset characteristics, validation strategies, performance metrics, and technical innovations. The findings reveal that the majority of studies focused on object detection and segmentation tasks, with limited use of external validation and publicly available datasets. Only a few works introduced technical novelty in architectural design. Data leakage due to improper dataset partitioning and augmentation practices was identified in several studies, raising concerns about model generalizability. Most models reported high performance using metrics such as accuracy, precision, and recall, but lacked justification for metric selection and reproducibility assurance. This review emphasizes the need for rigorous data management practices in dental AI research. Future studies should adopt patient-level data splitting, utilize external validation sets, and prioritize model interpretability and open science practices. The review serves as a critical resource for guiding the development of clinically robust CNN-based caries detection systems.