Development and Validation of an AI-Based Deep Learning System for Automated Assessment and Clinical Reporting of Root Canal Length, Curvature, and Morphology in Anterior and Premolar Teeth Using CBCT
试验速览
- 阶段
- 不适用
- 状态
- 尚未招募
- 入组人数
- 155
- 主要终点
- Tooth-level diagnostic accuracy and agreement of the AI-based system for root canal anatomical assessment
研究概览
简要总结
The primary aim of this study is to develop, evaluate, and validate a deep learning-based software system capable of generating automated, comprehensive clinical reports that detect, segment, and quantify root canal curvature, total tooth length, and morphological configurations in maxillary and mandibular anterior and premolar teeth using Cone-Beam Computed Tomography (CBCT) datasets
详细描述
Primary goal is to develop and validate the diagnostic performance metrics (e.g., Dice similarity coefficient, sensitivity, accuracy) of an artificial intelligence (AI)-driven tool; deep learning-based software system. That is capable of generating automated, comprehensive clinical reports based on: Automatic segmentation & measurements of tooth length, root canal curvature, as well as segmentation and classifying morphological configurations in maxillary & mandibular anterior & premolar teeth using Cone-Beam Computed Tomography (CBCT) datasets.
It involves a two-phase workflow:
- Development Phase: Deep learning models (like 3D U-Net architectures) are trained on a large, historically anonymized dataset of CBCT scans.
- Validation Phase: To validate the model's segmentations, measurements and processing time for the automated clinical reports against expert reference standards of manual measurements performed by expert endodontists and oral and maxillofacial radiologists
研究设计
- 研究类型
- Observational
- 观察模型
- Other
- 时间视角
- Retrospective
入排标准
- 性别
- All
- 接受健康志愿者
- 是
入选标准
- •Pre-existing CBCT DICOM datasets of maxillary & mandibular anterior & premolar teeth.
- •Permanent teeth with fully developed roots and completely formed apices.
- •High-quality CBCT scans with sufficient image resolution and minimal artifacts to permit accurate visualization of the root canal system.
- •Teeth with intact root canal anatomy suitable for automated analysis of root canal length, curvature, number of roots, number of root canals, and canal morphology.
- •Teeth representing a wide range of normal anatomical variations and root canal configurations to ensure adequate diversity for AI model development and validation.
- •DICOM datasets suitable for image preprocessing, annotation, and deep learning analysis.
排除标准
- •Teeth with incomplete root formation or open apices.
- •Teeth that have undergone previous endodontic treatment, retreatment, apexification, regenerative endodontic procedures, or root-end surgery.
- •Teeth with extensive coronal destruction, large restorations, metallic posts, intracanal filling materials, or crowns that obscure the root canal anatomy.
- •Teeth with root fractures, perforations, internal or external root resorption, severe root dilacerations, or other conditions that interfere with accurate anatomical assessment.
- •Teeth associated with extensive periapical lesions, cysts, tumors, or other pathological conditions that significantly alter the normal root canal anatomy.
- •Teeth with developmental anomalies affecting root morphology, including fusion, gemination, dens invaginatus, dens evaginatus, taurodontism,
- •CBCT datasets with severe motion artifacts, beam-hardening artifacts, metallic artifacts, excessive image noise, or poor image quality that compromises accurate image analysis.
- •Duplicate or incomplete DICOM datasets, corrupted image files, or datasets unsuitable for AI processing.
研究组 & 干预措施
Anonymized CBCT Datasets
Pre-existing anonymized CBCT DICOM datasets of maxillary and mandibular anterior and premolar teeth meeting the predefined eligibility criteria. The datasets will be used for development, validation, and independent testing of the deep learning-based system for automated assessment of root canal length, curvature, number of roots and canals, and root canal morphology. The independent validation set will include 155 selected teeth.
干预措施: AI-Based Deep Learning System for Root Canal Assessment (Diagnostic Test)
结局指标
主要结局
Tooth-level diagnostic accuracy and agreement of the AI-based system for root canal anatomical assessment
时间窗: through study completion, an average of 1 year
Diagnostic accuracy and agreement of the AI-based system will be assessed at the tooth level by comparing AI-generated measurements and classifications of root canal length, curvature, number of roots and canals, and canal morphology with the reference standard established by a panel of expert endodontists and oral and maxillofacial radiologists. Continuous and categorical outcomes will be evaluated using appropriate agreement and diagnostic accuracy measures.
次要结局
未报告次要终点
研究者
Eslam Fathy Bakhit Taha
Postgraduate student
Cairo University