Diagnostic Performance of Artificial Intelligence for Dental Caries Detection Across Clinical Imaging Modalities: Dataset for Systematic Review and Meta-analysis

This dataset accompanies the systematic review and meta-analysis titled “Artificial Intelligence for Dental Caries Detection Across Clinical Imaging: Diagnostic Performance and Implementation Barriers.” It contains all data extracted, curated, and analyzed during the study, including structured information on study characteristics, imaging modalities, AI model architectures, dataset properties, diagnostic accuracy metrics, and reported clinical implementation barriers.

The dataset was compiled following PRISMA 2020 recommendations and includes only human-derived clinical imaging studies. All variables were extracted independently by two reviewers, cross-validated, and harmonized using standardized coding procedures. For quantitative synthesis, the dataset provides complete 2×2 diagnostic contingency tables (TP, FP, FN, TN) required for sensitivity, specificity, DOR, and HSROC estimations. Studies lacking sufficient diagnostic cell counts are included in a separate sheet for transparency.

Additionally, the dataset contains a comprehensive matrix of the risk of bias assessment performed using an AI-adapted QUADAS-2 tool, as well as a structured corpus of reported implementation barriers. These textual data were used to generate the hierarchical semantic clustering model presented in the article.

The dataset is organized into the following files:

Study_Characteristics.xlsx – General study characteristics, imaging modalities, AI architectures, clinical setting, comparator, dataset features, and reference standard.

DTA_Contingency_Tables.xlsx – TP, FP, FN, TN values used for meta-analysis.

QUADAS2_AI_Risk_of_Bias.xlsx – Domain-level and overall risk-of-bias assessments.

Implementation_Barriers_Corpus.xlsx – Verbatim extracted barrier statements used for semantic clustering.

Search_Strategies.pdf – Complete reproducible search strategies for all databases.

Data_Dictionary.pdf – Variable definitions and coding conventions for reproducibility.

This dataset enables full transparency, reproducibility, and reusability of all analyses performed in the review. It may be used to replicate the meta-analytic models, evaluate diagnostic accuracy trends in AI-based caries detection, or support future evidence syntheses on artificial intelligence in dental imaging.