Diagnostic Accuracy of Artificial Intelligence to Predict the Need for Orthodontic Extraction Versus Non-Extraction- A Systematic Review and Meta-Analysis

Abstract

Objective:
To evaluate the diagnostic accuracy of artificial intelligence (AI) models in predicting the necessity for orthodontic extractions versus non-extractions.

Material and Methods – A systematic review was conducted following PRISMA-DTA guidelines, with studies selected based on the PIRD (Population, Index test, Reference standard, Diagnosis) framework. The inclusion criteria focused on diagnostic accuracy studies that used AI models to predict orthodontic extractions. Searches were performed across PubMed, DOAJ, EBSCO, K-hub, and Google Scholar for articles published from January 2000 to December 2023. 12 studies met the criteria, with a pooled meta-analysis involving four studies. Data were extracted and assessed for quality using QUADAS-2. Diagnostic accuracy metrics, including sensitivity, specificity, and receiver operating characteristic (ROC) curves, were analyzed using MetaDiSc version 1.4.

Results –The meta-analysis of 4 studies revealed a pooled sensitivity of 0.73 (95% CI: 0.70–0.76) and specificity of 0.82 (95% CI: 0.80–0.84), The ROC curve yielded an area under the curve (AUC) of 0.8813, Subgroup analysis highlighted that multi-layer perceptron (MLP) models had a sensitivity of 0.83 (95% CI: 0.79–0.87).

Conclusion - PEEK Retainers demonstrated superior Survival Rates and comparable Stability and Periodontal Health outcomes compared to Conventional Retainers. Although both materials performed acceptably, PEEK may offer enhanced durability, making it a viable alternative for long-term retention.