A new multi-modal dataset for Dental Plaque Diagnosis of Patients With Fixed Labial Orthodontic Appliances (Part 1 of 2)
Description
This dataset represents Part 1 of a two-part release of a multi-modal intraoral image dataset for dental plaque diagnosis in patients with fixed labial orthodontic appliances. The complete dataset is split into two parts due to repository size constraints. Part 2 (DOI: 10.17632/xjs4bfgzj5.2) contains the remaining patient data collected under the same clinical protocol, annotation scheme, and quality control procedures, and should be used jointly with this dataset for comprehensive benchmarking and analysis. Objective: This dataset is designed to benchmark and improve plaque diagnosis performance in patients wearing labial fixed orthodontic appliances. It addresses the lack of high-quality, large-scale intraoral image datasets collected from orthodontic patients, where brackets and wires increase diagnostic difficulty. Data content: The dataset includes two data types collected from 148 patients undergoing fixed orthodontic treatment. Intraoral images: * More than 10,000 pre-staining intraoral images spanning from the second premolar on one side to the contralateral second premolar (teeth 5–5), in both the maxillary and mandibular arches. * Images are annotated into four plaque severity levels with labels 0–1, 2, 3, and 4, representing increasing plaque accumulation. * Post-staining images were used as a reference for accurate labeling but are not included in the released dataset. Clinical records: * Each patient is associated with one clinical examination form. * Records include gingival bleeding assessment and oral hygiene habits, documented at both patient and tooth levels. Methodology: - Data collection: Images were captured using a professional dental camera following a standardized clinical protocol. For each patient, nine images were taken from nine predefined angles to comprehensively cover the anterior dentition. - Data labeling: All pre-staining images were classified into four plaque severity labels (0–1, 2, 3, 4). - Peer review: All annotations were independently reviewed by two dentists with over five years of clinical experience. - Noise creation: For each original image, three noise-augmented versions were generated using image shake, brightness increase, and brightness decrease. - Data augmentation: Six augmentation techniques were applied, resulting in 72 images per patient to enhance data diversity and robustness. Usage: The dataset is split into training and testing sets and is intended for training, fine-tuning, and evaluating deep learning models for plaque level diagnosis in patients wearing orthodontic braces.