Back to all papers

Artificial intelligence for upper-airway segmentation, adenoid-hypertrophy assessment, and sinonasal image analysis in orthodontic and craniofacial imaging: A systematic review and meta-analysis.

September 17, 2026pubmed logopapers

Authors

Guimarães HHDSD,Matos KN

Affiliations (2)

  • Department of Restorative Dentistry, Federal University of Minas Gerais, Minas Gerais, Brazil.
  • Department of Dentistry, São Paulo University, Avenida Professor Lineu Prestes, 2227, 05508-000 São Paulo, Brazil. Electronic address: [email protected].

Abstract

To systematically evaluate evidence on artificial intelligence (AI) for upper airway-related image analysis, with emphasis on upper-airway segmentation, adenoid-hypertrophy assessment, and sinonasal applications in orthodontic and craniofacial imaging. The protocol was registered in PROSPERO (CRD420261405535). Eligible studies assessed AI-based segmentation, detection, classification, volumetric assessment, software validation, or landmark localization using CBCT, CT, lateral cephalograms, panoramic radiographs, or CBCT-derived images. Data were synthesized qualitatively. The primary meta-analysis included studies reporting Dice similarity coefficient as mean and standard deviation for AI-based three-dimensional CBCT/CT upper or pharyngeal-airway segmentation, and a sensitivity analysis was performed in a more methodologically comparable subset. Risk of bias was assessed with QUADAS-2 and certainty of evidence with GRADE. Twenty-four studies were included across six domains. Seven studies were included in the primary meta-analysis, yielding a pooled Dice similarity coefficient of 0.935 (95% CI, 0.918-0.951; I<sup>2</sup>=48.6%). Sensitivity analysis restricted to four studies yielded a pooled Dice similarity coefficient of 0.956 (95% CI, 0.940-0.971; I<sup>2</sup>=28.5%). AI performance was generally favourable but varied according to imaging modality, anatomical target, model architecture, reference standard, population characteristics, and validation strategy. Evidence for adenoid-hypertrophy assessment was expanded by the inclusion of an additional CBCT-based nnU-Net study. External and multicentre validation remained limited, and certainty of evidence was moderate across domains. AI-based methods show promising performance, particularly for three-dimensional airway segmentation and emerging adenoid-hypertrophy applications. However, anatomical and methodological heterogeneity limits direct cross-study comparisons. Standardized anatomical definitions, representative populations, external validation, and clinically meaningful reference standards are needed before broad implementation.

Topics

Journal ArticleReview

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.