1. Background and Rationale The visual diagnosis of peri-implant mucosal erythema (redness), a key sign of inflammation, is highly subjective and varies significantly among clinicians, leading to inconsistencies in early detection and monitoring of peri-implant diseases. There is a critical need for an objective, quantitative, and reliable tool to standardize this assessment. Recent advances in artificial intelligence (AI) and colorimetric analysis of digital intraoral scans offer a promising solution to this clinical challenge. 2. Primary Objectives This diagnostic study aims to: Develop and validate a core colorimetric index that objectively quantifies mucosal erythema from digital intraoral scan data. Develop and validate an AI model that automatically calculates this index and provides a binary diagnosis (erythema present/absent) at the image level. Develop and validate a second AI model for precise localization (object detection) of erythematous regions on standard clinical software screenshots. Evaluate the clinical utility of the AI system by assessing its impact on the diagnostic accuracy, consistency, and confidence of clinicians with varying experience levels. 3. Study Design This is a multiphase diagnostic accuracy study conducted at a single academic center. It comprises three sequential phases with independent validation: Phase 1 (Development \& Internal Validation): Analysis of intraoral scans to derive the color index and train the AI models using an internal dataset. Phase 2 (External Technical Validation): Prospective validation of the trained AI models on an independent cohort of patients from a separate branch of the hospital. Phase 3 (Clinical Utility Assessment): A prospective, controlled, observer study where clinicians perform diagnoses with and without AI assistance. 4. Participants and Methods Data Source: Adult patients with dental implants who received intraoral scans using a 3Shape TRIOS 3 scanner. Image Data: Two formats are used: 1) Processed 3D surface files (PLY format) for colorimetric analysis, and 2) Standardized 2D screenshots from the 3Shape software for object detection. Reference Standards: Expert consensus on erythema (primary) and Bleeding on Probing (BOP, clinical inflammatory standard). AI Development: Deep learning models (e.g., convolutional neural networks) will be trained for index calculation, image-level diagnosis, and region localization. Observer Study: Participating clinicians (experts, general dentists, and students) will diagnose a set of test images both unaided and with AI assistance (which displays the color index value and/or bounding boxes). 5. Key Outcome Measures Diagnostic Accuracy: Area under the receiver operating characteristic curve (AUC), sensitivity, specificity (with 95% confidence intervals). Technical Performance: Intraclass correlation coefficient (ICC) for automated measurement agreement; Mean Average Precision (mAP) and Dice Similarity Coefficient for object detection. Clinical Impact: Change in diagnostic accuracy (AUC), inter-observer agreement (Kappa), and diagnostic confidence scores when using AI assistance. 6. Significance This study seeks to translate a subjective clinical sign into an objective, AI-powered diagnostic biomarker. If successful, the proposed system could become a valuable decision-support tool in daily practice and clinical research, promoting earlier, more consistent, and standardized monitoring of peri-implant tissue health, ultimately improving patient care.
Study Type
INTERVENTIONAL
Allocation
NA
Purpose
DIAGNOSTIC
Masking
NONE
Enrollment
200
Participants in this single-arm study undergo evaluation using the investigational AI-based colorimetric system. The study involves two distinct participant roles: 1) Patient Participants who have previously received intraoral scans contribute their de-identified digital dental images (3D surface files and 2D screenshots) for AI model development and validation. 2) Clinician Participants (including experts, general dentists, and students) take part in a prospective observer study. In a controlled, crossover manner, they diagnose a standardized set of peri-implant mucosal images first without any aid, and then with the assistance of the AI system, which provides an objective color index value and visual bounding boxes around suspected erythematous regions. The primary aim for this arm is to assess the diagnostic accuracy, reliability, and clinical utility of the AI system across both technical (vs. expert reference) and human (clinician performance enhancement) endpoints.
Department of Oral Maxillofacial Implantology Shanghai Ninth People's Hospital
Shanghai, China
RECRUITINGDiagnostic accuracy of the AI system for detecting peri-implant mucosal erythema, as measured by the Area Under the Receiver Operating Characteristic Curve (AUC).
The primary outcome is the diagnostic accuracy of the AI-based colorimetric system in classifying an image as showing erythema or not. Accuracy is quantified by the Area Under the Receiver Operating Characteristic Curve (AUC), with expert visual diagnosis serving as the reference standard. The AUC, along with its 95% confidence interval, will be calculated separately for the internal development set and the independent external validation set to assess model performance and generalizability.
Time frame: At the completion of image analysis for the external validation set, approximately 3 months after study start
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.