Scholarlink

<p><b>Objectives</b> This study aimed to collect, categorise, and annotate a comprehensive dataset of intra-oral clinical images specifically designed for artificial intelligence training and testing.</p><p><b>Materials and methods</b> Full-mouth clinical photos were collected from patients attending the Oral Medicine Clinic at the Faculty of Dentistry, Cairo University. A clinical oral examination was performed by two oral medicine specialists to establish and document the clinical diagnosis. The dataset was comprehensively annotated using LabelMe.exe in JavaScript Object Notation (JSON) format.</p><p><b>Results</b> The dataset comprises 9,201 intra-oral images, which are subdivided according to the presence and type of oral lesion. These include 4,405 images classified as ‘normal', 2,314 as ‘low risk', and 2,482 as ‘high risk'. The image dimensions range from a minimum of 40,992 pixels to a maximum of 24,216,480 pixels. It includes a wide variety of oral lesions in all sites of the oral cavity, ensuring a comprehensive representation of different diseases. A significant number of images present periodontal diseases, while the dataset also features various classes of carious lesions from different intra-oral views, supporting research in conservative dentistry.</p><p><b>Conclusion</b> Researchers can use the annotated dataset in the JSON format for training, validating and testing deep learning algorithms.</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

An annotated clinical image dataset for AI classification of malignant and potentially malignant oral lesions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月12日

查看全文
  • Noran Ayman,
  • Fat´heya M. Zahran,
  • Yomna Safaa El-Din,
  • Noha Adel Azab

摘要

Objectives This study aimed to collect, categorise, and annotate a comprehensive dataset of intra-oral clinical images specifically designed for artificial intelligence training and testing.

Materials and methods Full-mouth clinical photos were collected from patients attending the Oral Medicine Clinic at the Faculty of Dentistry, Cairo University. A clinical oral examination was performed by two oral medicine specialists to establish and document the clinical diagnosis. The dataset was comprehensively annotated using LabelMe.exe in JavaScript Object Notation (JSON) format.

Results The dataset comprises 9,201 intra-oral images, which are subdivided according to the presence and type of oral lesion. These include 4,405 images classified as ‘normal', 2,314 as ‘low risk', and 2,482 as ‘high risk'. The image dimensions range from a minimum of 40,992 pixels to a maximum of 24,216,480 pixels. It includes a wide variety of oral lesions in all sites of the oral cavity, ensuring a comprehensive representation of different diseases. A significant number of images present periodontal diseases, while the dataset also features various classes of carious lesions from different intra-oral views, supporting research in conservative dentistry.

Conclusion Researchers can use the annotated dataset in the JSON format for training, validating and testing deep learning algorithms.