Root canal preparation in endodontics poses significant challenges, particularly in curved canals of mandibular molars, where accurate preoperative assessment using CBCT imaging is crucial to avoid iatrogenic errors and improve treatment outcomes. This study aims to develop and evaluate the diagnostic accuracy of a deep learning model for analyzing root canal curvature angles in mandibular molars from CBCT scans, compared to human expert evaluations. The model will leverage advanced AI techniques to segment and measure curvatures objectively, addressing limitations in manual interpretation, potentially standardizing case difficulty assessments and aiding clinical decision-making.

Inclusion Criteria:

• CBCT scans of mandibular molars of Egyptian patients aging from 18 to 65 years old
• Small Field of view (FOV) including maximum a quadrant
• Voxel size not larger than 2mm
• Mandibular molars showing complete root formation
• Carious or non-carious teeth
• Absence of artifacts, dental implants in the adjacent teeth

Exclusion Criteria:

• Mandibular first and second molars with developmental anomalies, external or internal root resorption, root canal calcification, previous root canal treatment, post restorations, and/or root caries
• CBCT images of sub-optimal quality or artifacts/high scatter interfering with proper assessment

The accepted sample size (n) will be 207, when the accuracy is (90.48%) and marginal error (4 %) by adopting confidence interval (95%), using Staskingdom calculator.

The study is diagnostic accuracy study on CBCT data