Comparison of cone beam computed tomography measurements obtained after manual and AI-assisted head positioning and reconstruction.
Authors
Affiliations (3)
Affiliations (3)
- Division of Oral Radiology, Faculdade São Leopoldo Mandic, Instituto de Pesquisas São Leopoldo Mandic, Campinas, SP, Brazil.
- Department of Clinical Dentistry, Federal University of Espírito Santo (UFES), Espírito Santo, Brazil. [email protected].
- Department of Oral Diagnosis, Piracicaba Dental School, University of Campinas, Piracicaba, SP, Brazil.
Abstract
To compare measurements obtained from cone beam computed tomography (CBCT) exams by radiologists following manual head positioning and panoramic curve tracing with those obtained after artificial intelligence (AI)-assisted positioning adjustments performed prior to measurement. Twenty-five CBCT exams presenting edentulous sites corresponding to missing mandibular first molars were selected. Five examiners performed measurements of bone height and width after manual head positioning adjustments and panoramic curve tracing (manual method). Subsequently, the same exams were processed using TomoNav software, which automatically performed volume alignment and generated the panoramic curve and cross-sectional images. The examiners then repeated the measurements (AI-assisted method). Measurements obtained from both approaches were summarized using descriptive statistics and compared using the intraclass correlation coefficient (ICC), paired t-test, and Bland-Altman analysis (p˂0.05). Excellent inter-examiner reproducibility was observed for both height and width measurements, with ICC values of 0.94 for height and 0.81 and 0.85 for width using the manual and AI-assisted methods, respectively. Significant differences were identified for left-side height, total height, right-side width, and total width (p < 0.05). However, the mean biases were low and overall agreement between methods was excellent, with ICC values equal or greater than 0.89. Measurements obtained with the AI-assisted method for volume repositioning and generation of panoramic and cross-sectional images were comparable to those obtained using a conventional method. AI facilitates CBCT implant-planning preprocessing without compromising measurement reproducibility.