Mandibular Third Molar (m-TM) Dataset

Overview

The m-TM dataset was created to address the lack of publicly available, labeled datasets specifically designed for detecting mandibular third molars (m-M3). Derived from the UESB dataset (Silva et al., 2018), which originally consisted of 1500 panoramic radiographs (PRs) intended for tooth segmentation tasks, the m-TM dataset has been repurposed into a specialized resource for m-M3 presence detection. During the preparation process, 183 PRs were excluded due to factors such as the m-M3 being in the germination stage, misalignment caused by missing teeth, anomalies within the mandibular region, or inadequate image quality. The remaining 1317 PR images were meticulously labeled by an expert oral and maxillofacial radiologist, creating a problem-specific, high-quality dataset that supports advanced research in AI-assisted dentistry.

Dataset Details

The m-TM dataset is organized into four distinct categories based on the presence or absence of m-M3 in the left and right mandibular regions. It includes 89 images where m-M3 is absent on the left but present on the right, 74 images where m-M3 is present on the left but absent on the right, 810 images where m-M3 is present on both sides, and 344 images where m-M3 is absent on both sides. This structured classification enhances the dataset's utility for benchmarking and facilitates advanced AI-driven dental research. Additionally, the dataset adheres to the FAIR Guiding Principles, ensuring it is findable, accessible, interoperable, and reusable. These principles support the dataset's accessibility and promote reproducibility, making it a valuable resource for the research community.

Significance

The m-TM dataset represents a pioneering contribution to AI-assisted dental research, serving as the first publicly available dataset explicitly developed for m-M3 presence detection. By adhering to the FAIR principles and incorporating expert annotations, the dataset establishes a standardized benchmark for researchers, facilitates interdisciplinary collaboration, and promotes the development of advanced AI methodologies for dental applications. This resource fills a critical gap in the field, supporting innovations in clinical decision support systems and advancing the integration of AI into dentistry.

Citations

Researchers utilizing the m-TM dataset are required to cite the following studies:

- Kayadibi, İ., Köse, U., Güraksın, G. E., & Çetin, B. (2025). An AI-assisted explainable mTMCNN architecture for detection of mandibular third molar presence from panoramic radiography. International Journal of Medical Informatics, 195, 105724.

- Silva, G., Oliveira, L., & Pithon, M. (2018). Automatic segmenting teeth in X-ray images: Trends, a novel dataset, benchmarking, and future perspectives. Expert Systems with Applications, 107, 15–31.

For additional details about the UESB dataset, please visit the following website: http://ivisionlab.ufba.br/