Expert Annotated Mandibular Third Molar (ExAn-MTM) Dataset

📌 Overview of ExAn-MTM Dataset

The ExAn-MTM dataset was developed to address the scarcity of publicly available annotated datasets for mandibular third molar (MTM) detection. It is derived from high-quality panoramic radiographs (PRs) and consists of 973 expertly annotated images, adapted from the original m-TM dataset.

This dataset provides a reliable benchmark for AI-driven dental diagnostics, computer vision research, and medical imaging studies.

📂Dataset Structure- The dataset includes bounding box annotations for two clinically important classes of mandibular third molars:

0 → e-MTM (erupted mandibular third molar)

1 → i-MTM (impacted mandibular third molar)

Annotations were created by an experienced oral and maxillofacial radiologist using the MakeSense annotation tool, and are provided in YOLO format.

Folder structure:

/ExAn-MTM dataset/
├─ train/
│ ├─ images/
│ ├─ labels/
├─ valid/
│ ├─ images/
│ ├─ labels/

Data split:

Training Set → 875 images (624 e-MTM, 980 i-MTM)

Validation Set → 98 images (63 e-MTM, 117 i-MTM)

🎯 Applications

The dataset can be used for:

Training and validating object detection models (YOLO, Faster R-CNN, RetinaNet, etc.)

Benchmarking AI-based clinical decision support systems (CDSS) in dentistry

Conducting comparative studies in medical imaging and annotation methods

Advancing research in explainable AI (XAI) for dental applications

🌍 Significance

First publicly available, expert-annotated dataset dedicated to mandibular third molar detection.

Adheres to FAIR principles (Findable, Accessible, Interoperable, Reusable).

Enhances reproducibility and transparency in dental AI research.

Provides a solid foundation for developing and benchmarking diagnostic models in dental radiology.

📑 Citation

This dataset is derived from and must be cited alongside the following studies:

1- Kayadibi, İ., Köse, U., Güraksın, G.E. et al. E-MTMYOLO: an explainable YOLOv5-based architecture for accurate detection of mandibular third molar using a novel expert-annotated dataset. J Supercomput 81, 1286 (2025). https://doi.org/10.1007/s11227-025-07775-w

2- 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. https://doi.org/10.1016/j.ijmedinf.2024.105724