Lesion_Dataset: Periapical Radiographs for Lesion Classification

Lesion_Dataset is a collection of grayscale periapical radiographic images in TIFF (.TIF) format. Expert clinicians curated the images from routine dental examinations of 1,295 patients conducted at a dental clinic in Ankara, Türkiye, between 2006 and 2023. Each image corresponds to a distinct periapical view of a tooth root.

Labels: image-level binary labels based on expert visual assessment only (no pixel-wise annotations or segmentation masks).

Lesioned: 1,192 images

Healthy: 1,093 images

Total: 2,285 images

File naming convention: files are named as PXXXXXX_YYY.TIF. The prefix before the underscore denotes a pseudonymous patient identifier (e.g., P000824). Files sharing the same prefix belong to the same patient.

License: Creative Commons Attribution 4.0 International (CC BY 4.0).

Suggested citation: https://doi.org/10.5281/zenodo.18824283

Related publications (please cite if you use the dataset in academic work):

Can, Z., & Aydin, E. (2025). Explainable CNN–Radiomics Fusion and Ensemble Learning for Multimodal Lesion Classification in Dental Radiographs. Diagnostics, 15(16), 1997.

Can, Z., Isik, S., & Anagun, Y. (2024). CVApool: Using the null-space of CNN weights for tooth disease classification. Neural Computing and Applications, 36(26), 16567–16579.

Contact: zcan@ogu.edu.tr
 | https://web.ogu.edu.tr/zuhalcan/