Performance of artificial intelligence using cone-beam computed tomography for segmentation of oral and maxillofacial structures: A systematic review and meta-analysis.

ABSTRACT

Background: There are different values reported about the performance of artificial intelligence using cone-beam computed tomography (CBCT) for segmentation of oral and maxillofacial structures. We aimed to perform a systematic review and meta-analysis to provide an overall estimate to resolve the given conflicts. Material and Methods: A literature search was conducted in Embase, PubMed, and Scopus through 31 October 2022, to identify studies evaluating artificial intelligence systems using oral and maxillofacial CBCT images for automatic segmentation of anatomical landmarks. The surveys had to report the outcome according to dice coefficient (DICE) or dice similarity coefficient (DSC) index. The estimates were presented as percent and 95% confidence interval (CI). I-squared index was used to assess the heterogeneity between the surveys. Results: A total of 24 eligible studies were finally enrolled. The overall pooled DICE/DSC value for artificial intelligence was 0.92 (95% CI 0.88-0.95; I-squared=93.6%, p<0.001). Tooth and mandible were evaluated more than other anatomical regions (five studies for each one). The lowest and highest DICE/DSC scores for the artificial intelligence related to inferior alveolar nerve (0.55 [95% CI 0.47-0.63]) and pharyngeal airway and sinonasal cavity (0.98 [95% CI 0.98-1.00]). Conclusions: The findings revealed excellent performance for the artificial intelligence regarding the segmentation task of oral and maxillofacial CBCT images. Key wordsArtificial intelligence, cone-beam computed tomography, segmentation performance, dentistry.