Artificial Intelligence–Driven Alerts to Detect Poor Oral Hygiene and Appliance Misuse in Orthodontic Patients: A Feasibility Study

Abstract

Orthodontic treatment outcomes depend heavily on patient cooperation and oral hygiene, yet monitoring adherence remains a major challenge in clinical practice. This retrospective feasibility study evaluated an artificial intelligence (AI)–driven alert system designed to detect poor oral hygiene and appliance misuse in orthodontic patients. A total of 260 patients treated with either fixed appliances (n = 160) or aligners (n = 100) between January 2018 and January 2019 were included. Archived intraoral photographs, clinical notes, and aligner usage logs were analyzed using a two-module AI system: a convolutional neural network for plaque and hygiene detection, and an object recognition module for appliance misuse. Diagnostic accuracy was assessed against orthodontist documentation. The AI system demonstrated high diagnostic performance, with overall sensitivity and specificity of 86.7% and 82.0% for hygiene detection, and 81.6% and 83.1% for appliance misuse, respectively. Agreement with clinician documentation was substantial (κ = 0.67–0.73). Younger patients generated the highest frequency of alerts, particularly in the fixed appliance group, reflecting greater challenges with compliance. On average, analysis time was less than three minutes per case, and only 12.3% of alerts required orthodontist override. Operational feasibility was supported by simulated integration into electronic health records and high inferred patient acceptability. These findings suggest that AI-driven alerts can provide accurate, efficient, and clinically relevant monitoring of orthodontic patients. By enabling early detection of hygiene lapses and misuse, such systems hold promise for reducing preventable complications, enhancing patient accountability, and supporting a proactive model of orthodontic care. Prospective trials are warranted to validate effectiveness in real-world practice.