SMART Intraoral images Dataset

About Dataset

The SMART (Systematic Oral Mucosal Annotated Images for Research and Training) dataset has been developed as part of the Indian Council of Medical Research (ICMR)-funded project titled “Development of a Simple Mobile Interface Technology Application (SMITA) for the Screening of Oral Disorders Among High-Risk Rural Populations in Tamil Nadu” (Project ID: IIRP–2023–1049). This initiative aimed at enabling advancements in mobile-based Oral Health screening, clinical education, and Artificial Intelligence applications in Dentistry. The dataset comprises 1850 intraoral images collected from 235 patients attending various spokes who have been identified as a part of this project. The dataset will be updated monthly.

For each patient, a standardized protocol was followed to capture eight intraoral photographs covering key anatomical sites: dorsal tongue, ventral tongue, left buccal mucosa, right buccal mucosa, upper lip, lower lip, upper dental arch, and lower dental arch based on the manual developed and copyrighted as a part of this project. These images were obtained using mobile phone cameras under field conditions, ensuring practical applicability in low-resource or community-based settings, captured by Healthcare professionals.

The dataset is categorized into four diagnostic classes: Normal mucosa, Variations from Normal, and Oral Potentially Malignant Disorders (OPMDs). Each image was meticulously annotated using the VGG Image Annotator (VIA) tool, with region-specific labelling based on a customized set of clinically relevant descriptors to maintain annotation uniformity and diagnostic precision.

The SMART dataset is intended to support AI model development for diagnostic decision-making, real-time image triaging, and educational training modules. All procedures were carried out with appropriate institutional ethics approvals, ensuring strict adherence to data privacy, participant confidentiality, and responsible data sharing practices.