Smartphone Images for Multi-Class Dental Caries Detection: A Pilot Study

A dataset of six self-taken intraoral smartphone images from seventy adults, annotated by four dentists for dental caries detection. The images were labeled using binary, four-class, and five-class schemes based on ICDAS-II and ADA CCS criteria. The dataset was used to train a YOLOv6-L6 deep learning model, with performance metrics reported.

Use Cases

• Training deep learning models for automated dental caries detection based on smartphone images.
• Benchmarking model performance on multi-class classification tasks for dental health.
• Studying the feasibility of patient-driven screening tools based on self-captured images.
• Analyzing inter- and intra-examiner reliability in dental image annotation.

Strengths

• Images were annotated by four dentists with substantial intra-examiner reliability (κ = 0.765–0.952).
• Model performance metrics are reported, including sensitivity ≥86% and specificity >97% across classification schemes.
• Dataset is openly available under a CC-BY-4.0 license.

Limitations

• Dataset scale is limited to a pilot study with 70 participants and 32.3 KB total size.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Ji-Yun Kang via figshare

Collection Method

Images were self-taken by seventy adults in a non-clinical setting and annotated by four dentists.

Time Range

null

Freshness

Last updated 2026-05-08 05:51:45; freshness should be verified.

Geography

null