Non-Cochrane · DTA · PM40802701
Examining the diagnostic accuracy of artificial intelligence for detecting dental caries across a range of imaging modalities: An umbrella review with meta-analysis.
Arzani Sarah, Karimi Ali, Iranmanesh Pedram, et al. · PloS one · 2025 · PubMed · doi:10.1371/journal.pone.0329986
Full textEvery question was extracted from the full text, figures and supplements by two independent LLM extractors, disagreements adjudicated. Rows cite where in the review the value comes from.
Questions
26
current version
GRADE rated
0
No studies
0
Included studies
20
linked below
No studies 0 · 0%Ungraded 26 · 100%Very low 0 · 0%Low 0 · 0%Moderate 0 · 0%High 0 · 0%
Ungraded means an effect was reported without a GRADE rating; it is not a gap. No studies means the review found nothing to answer the question. The six categories are a display order, not a score.
Review question
| Element | Planned (methods) | Found (results) |
|---|---|---|
| Population | Systematic reviews/meta-analyses evaluating diagnostic accuracy of AI algorithms for dental caries detection (PIRD; human subjects, relevant dental images) | 14 systematic reviews included (12 SR, 2 SR with meta-analysis) spanning 137 primary studies (originals 1984-2023); 20 primary DTA studies pooled (29423 diagnostic tests; caries prevalence 27.3%); moderate overlap (CCA 6.90%) |
| Intervention | AI algorithms for dental caries detection (index test; e.g., CNN, ANN, SVM, deep/machine learning) | AI algorithms, predominantly CNN (ResNet, Faster R-CNN, U-net, VGG16/19, Inception V3, MobileNet, DenseNet, YOLOv3, EfficientNet, etc.) |
| Comparison | Visual-tactile or clinical-based assessment of dental caries (reference standard) | Dental expert / visual-tactile clinical assessment as reference standard |
| Outcomes | Diagnostic accuracy: sensitivity, specificity, log diagnostic odds ratio, area under ROC curve (AUC) | Pooled sensitivity 0.85 (0.83-0.93), specificity 0.90 (0.85-0.95), log DOR 4.37 (3.16-6.27) [DOR 70.9, 44.9-111.9], AUC 0.86; LR+ 10.443 (7.505-14.531), LR- 0.168 (0.138-0.205); positive/negative post-test probability 79%/6%; I2>95%; publication bias present (Egger/Begg p<0.05; trim-and-fill imputed 6 studies for sensitivity and log DOR) |
| Study designs | Dental imaging across modalities (intraoral photographs, periapical, bitewing, panoramic radiographs, CBCT, near-infrared transillumination) | Imaging modalities: intraoral photographs, periapical, bitewing, panoramic radiographs, CBCT, NILT |
| Timing | Cross-sectional diagnostic accuracy (index test vs reference standard) | Cross-sectional diagnostic accuracy; included primary studies published 1984-2023 |
Search date: October 15, 2024. Databases: Cochrane Library; Embase; PubMed; Scopus; Web of Science; WorldCat; Google Scholar (first 100 hits, grey literature)
Questions and results
AI algorithms (predominantly CNN/deep learning) as index test vs Dental expert / visual-tactile clinical assessment (reference standard)
Population: Patients undergoing dental caries detection on dental images (teeth/surfaces across imaging modalities)
| Outcome | Effect | Evidence | Status / GRADE |
|---|---|---|---|
Sensitivity Cross-sectional diagnostic accuracy (index test vs reference standard) · Overall (all imaging modalities pooled) AI algorithms have high pooled sensitivity (0.85) for detecting dental caries; ungraded, with very high heterogeneity (I2>95%) and evidence of publication bias. Source: Abstract; Results 'Meta-analysis of eligible original studies'; Fig 4; Table 6 · same question in other reviews | Sensitivity 0.85 [0.83, 0.93] | 20 studies · 29423 participants | Ungraded |
Specificity Cross-sectional diagnostic accuracy (index test vs reference standard) · Overall (all imaging modalities pooled) AI algorithms have high pooled specificity (0.90) for detecting dental caries; ungraded, with very high heterogeneity (I2>95%). Source: Abstract; Results 'Meta-analysis of eligible original studies'; Fig 4; Table 6 · same question in other reviews | Specificity 0.90 [0.85, 0.95] | 20 studies · 29423 participants | Ungraded |
Negative post-test probability Overall (all imaging modalities pooled) Population: Dental images or diagnostic tests for dental caries detection The negative post-test probability was 6%; certainty was not graded. Source: Results, 'Meta-analysis of eligible original studies' · same question in other reviews | Post-test probability 0.06 | 20 studies | Ungraded |
Area under the summary ROC curve Cross-sectional diagnostic accuracy (index test vs reference standard) · Overall (all imaging modalities pooled) The area under the summary ROC curve for AI caries detection is 0.86, indicating high overall diagnostic accuracy; ungraded. Source: Abstract; Results 'Meta-analysis of eligible original studies'; Fig 8; Fig 11 · same question in other reviews | AUC 0.86 | 20 studies · 29423 participants | Ungraded |
Positive post-test probability Overall (all imaging modalities pooled) Population: Dental images or diagnostic tests for dental caries detection The positive post-test probability was 79%; certainty was not graded. Source: Results, 'Meta-analysis of eligible original studies' · same question in other reviews | Post-test probability 0.79 | 20 studies | Ungraded |
Positive likelihood ratio Cross-sectional diagnostic accuracy (index test vs reference standard) · Overall (all imaging modalities pooled) A positive AI test raises caries probability substantially (LR+ 10.44; positive post-test probability 79%); ungraded. Source: Results 'Meta-analysis of eligible original studies'; Discussion; Fig 9 · same question in other reviews | LR+ 10.443 [7.505, 14.531] | 20 studies · 29423 participants | Ungraded |
Log diagnostic odds ratio Cross-sectional diagnostic accuracy (index test vs reference standard) · Overall (all imaging modalities pooled) AI algorithms show strong pooled discriminative accuracy for dental caries (log DOR 4.37; DOR 70.9); ungraded. Source: Abstract; Results 'Meta-analysis of eligible original studies'; Fig 4; Table 6 · same question in other reviews | Log DOR 4.37 [3.16, 6.27] | 20 studies · 29423 participants | Ungraded |
Negative likelihood ratio Cross-sectional diagnostic accuracy (index test vs reference standard) · Overall (all imaging modalities pooled) A negative AI test lowers caries probability substantially (LR- 0.17; negative post-test probability 6%); ungraded. Source: Results 'Meta-analysis of eligible original studies'; Discussion; Fig 9 · same question in other reviews | LR- 0.168 [0.138, 0.205] | 20 studies · 29423 participants | Ungraded |
Diagnostic odds ratio Overall (all imaging modalities pooled) Population: Dental images or diagnostic tests for dental caries detection The pooled diagnostic odds ratio was 70.9 (95% CI 44.9 to 111.9) across 20 studies; certainty was not graded. Source: Results, 'Meta-analysis of eligible original studies' · same question in other reviews | DOR 70.9 [44.9, 111.9] | 20 studies | Ungraded |
Comparison
| Outcome | Effect | Evidence | Status / GRADE |
|---|---|---|---|
Log diagnostic odds ratio Imaging modality: Bitewing radiographs | Log DOR 3.96 [3.20, 4.72] | 5 studies | Ungraded |
Specificity Imaging modality: Periapical radiographs | Specificity 0.88 [0.82, 0.93] | 12 studies | Ungraded |
Sensitivity (trim-and-fill adjusted for publication bias) Overall; trim-and-fill (20 observed + 6 imputed studies) | Sensitivity 0.818 [0.778, 0.858] | 26 studies | Ungraded |
Log diagnostic odds ratio Imaging modality: Intraoral photographic images | Log DOR 3.69 [2.80, 4.59] | 7 studies | Ungraded |
Sensitivity Imaging modality: CBCT | Sensitivity 0.73 [0.68, 0.77] | 1 study | Ungraded |
Sensitivity Imaging modality: Intraoral photographic images | Sensitivity 0.80 [0.69, 0.90] | 7 studies | Ungraded |
Log diagnostic odds ratio (trim-and-fill adjusted for publication bias) Overall; trim-and-fill (20 observed + 6 imputed studies) | Log DOR 3.913 [3.219, 4.606] | 26 studies | Ungraded |
Specificity Imaging modality: Panoramic radiographs | Specificity 0.97 [0.96, 0.98] | 8 studies | Ungraded |
Log diagnostic odds ratio Imaging modality: CBCT | Log DOR 6.33 [5.95, 6.71] | 1 study | Ungraded |
Log diagnostic odds ratio Imaging modality: Panoramic radiographs | Log DOR 5.08 [4.15, 6.02] | 8 studies | Ungraded |
Log diagnostic odds ratio Imaging modality: Periapical radiographs | Log DOR 4.72 [3.16, 6.27] | 12 studies | Ungraded |
Specificity Imaging modality: Intraoral photographic images | Specificity 0.84 [0.77, 0.91] | 7 studies | Ungraded |
Sensitivity Imaging modality: Panoramic radiographs | Sensitivity 0.83 [0.72, 0.94] | 8 studies | Ungraded |
Specificity Imaging modality: CBCT | Specificity 1.00 [0.99, 1.00] | 1 study | Ungraded |
Sensitivity Imaging modality: Periapical radiographs | Sensitivity 0.86 [0.83, 0.90] | 12 studies | Ungraded |
Sensitivity Imaging modality: Bitewing radiographs | Sensitivity 0.81 [0.72, 0.91] | 5 studies | Ungraded |
Specificity Imaging modality: Bitewing radiographs | Specificity 0.89 [0.84, 0.95] | 5 studies | Ungraded |
Included primary studies 20
Open as a trial list| Study | Also included by | Outcomes it informs here | Trial PICO | Results |
|---|---|---|---|---|
| Bayraktar Y 2022 Diagnosis of interproximal caries lesions with deep convolutional neural network in digital bitewing radiographs. PMID:34173051 | 1 other review | Sensitivity | not extracted yet | not extracted yet |
| De Araujo Faria V 2021 Prediction of Radiation-Related Dental Caries Through PyRadiomics Features and Artificial Neural Network on Panoramic Radiography. PMID:34254199 | 1 other review | Sensitivity | not extracted yet | not extracted yet |
| Devlin H 2021 The ADEPT study: a comparative study of dentists' ability to detect enamel-only proximal caries in bitewing radiographs with and without… PMID:34686815 | 1 other review | Sensitivity | not extracted yet | not extracted yet |
| Kühnisch J 2022 Caries Detection on Intraoral Images Using Artificial Intelligence. PMID:34416824 | 1 other review | Sensitivity | not extracted yet | not extracted yet |
| Zheng L 2021 Artificial intelligence in digital cariology: a new tool for the diagnosis of deep caries and pulpitis using convolutional neural networks. PMID:34268376 | 1 other review | Sensitivity | not extracted yet | not extracted yet |
| Ari T 2022 Automatic Feature Segmentation in Dental Periapical Radiographs. PMID:36553088 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Chen X 2022 Detection of Proximal Caries Lesions on Bitewing Radiographs Using Deep Learning Method. PMID:36215971 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Estai M 2022 Evaluation of a deep learning system for automatic detection of proximal surface dental caries on bitewing radiographs DOI:10.1016/j.oooo.2022.03.008 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Ezhov M 2021 Development and Validation of a Cbct-Based Artificial Intelligence System for Accurate Diagnoses of Dental Diseases DOI:10.21203/rs.3.rs-303329/v1 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Fariza A 2022 Evaluation of Convolutional Neural Network for Automatic Caries Detection in Digital Radiograph Panoramic on Small Dataset DOI:10.1109/icodse56892.2022.9972183 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Imak A 2022 Dental Caries Detection Using Score-Based Multi-Input Deep Convolutional Neural Network DOI:10.1109/access.2022.3150358 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Jayasinghe H 2022 Effectiveness of Using Radiology Images and Mask R-CNN for Stomatology DOI:10.1109/icac57685.2022.10025034 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Li S 2022 Artificial intelligence for caries and periapical periodontitis detection DOI:10.1016/j.jdent.2022.104107 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Liu F 2022 Recognition of Digital Dental X-ray Images Using a Convolutional Neural Network. PMID:36109403 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Liu L 2015 A Smart Dental Health-IoT Platform Based on Intelligent Hardware, Deep Learning, and Mobile Terminal DOI:10.1109/jbhi.2019.2919916 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Oztekin F 2023 An Explainable Deep Learning Model to Prediction Dental Caries Using Panoramic Radiograph Images. PMID:36673036 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Park EY 2022 Caries detection with tooth surface segmentation on intraoral photographic images using deep learning. PMID:36476359 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Suttapak W 2022 A unified convolution neural network for dental caries classification DOI:10.37936/ecticit.2022162.245901 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Vinayahalingam S 2021 The Automatic Detection of Caries in Third Molars on Panoramic Radiographs Using Deep Learning: A Pilot Study DOI:10.21203/rs.3.rs-379636/v1 | this review only | Sensitivity | not extracted yet | not extracted yet |
| Zadrożny Ł 2022 Artificial Intelligence Application in Assessment of Panoramic Radiographs. PMID:35054390 | this review only | Sensitivity | not extracted yet | not extracted yet |
Overlap with other reviews
Reviews that include the same primary studies are not independent evidence. Shared counts match studies by PMID, DOI or NCT, and otherwise by author and year (a probable match).
| Review | Shared studies | Jaccard |
|---|---|---|
| Accuracy of artificial intelligence in caries detection: a systematic review and meta-analysis. (2025) | 4 | 0.08 |
| Uses of Different Machine Learning Algorithms for Diagnosis of Dental Caries. (2022) | 1 | 0.03 |