Advanced proximal carious lesions · Permanent molar surfaces in diagnostic accuracy studies (mostly in vitro; clinical)
Pooled sensitivity for radiographic detection of advanced proximal lesions was 0.444 (95% CI 0.38 to 0.51).
8 findings from 1 review
Sensitivity 44.4 [38, 51]
22 studies
GRADE not reported
Negative LR 0.58 [0.52, 0.65]
22 studies
GRADE not reported
Positive LR 9.82 [7.3, 13.3]
22 studies
GRADE not reported
Show 5 more findings
Specificity 95.5 [94, 97]
22 studies
GRADE not reported
Negative LR 0.58 [0.52, 0.65]
22 studies
GRADE not reported
Specificity 95.5 [94, 97]
22 studies
GRADE not reported
Sensitivity 44.4 [38, 51]
22 studies
GRADE not reported
Positive LR 9.82 [7.3, 13.3]
22 studies
GRADE not reported
All proximal carious lesions · Permanent molar surfaces in diagnostic accuracy studies (mostly in vitro; clinical)
Pooled specificity for radiographic detection of all proximal lesions was 0.900 (95% CI 0.87 to 0.92).
1 finding from 1 review
Specificity · Patients undergoing dental X-ray (radiographic) imaging for caries detection
Across 7 studies, AI showed high weighted-mean specificity (~87.9%) for caries detection, with very high heterogeneity (I2=96%); ungraded.
1 finding from 1 review
Diagnostic accuracy (prevalence of accuracy) · Patients undergoing dental X-ray (radiographic) imaging for caries detection
On digital bitewing X-rays, AI accuracy was around 80% (single study, narrow CI; ungraded subgroup, numeric estimate not precisely readable).
4 findings from 1 review
Accuracy of artificial intelligence in caries detection: a systematic review and meta-analysis. 2025 All outcomesDental radiography (digital periapical, bitewing, panoramic; OCT/micro-CT in one study); studies 2017-2023 across China, South Korea, Brazil, Turkey, France, USA, UK, Taiwan, Germany, India, Netherlands · Digital bitewing X-rays (Fig 4a)
~80
1 study
GRADE not reported
Accuracy of artificial intelligence in caries detection: a systematic review and meta-analysis. 2025 All outcomesDental radiography (digital periapical, bitewing, panoramic; OCT/micro-CT in one study); studies 2017-2023 across China, South Korea, Brazil, Turkey, France, USA, UK, Taiwan, Germany, India, Netherlands · Periapical X-rays (Fig 4c)
~76
Studies not reported
GRADE not reported
Accuracy of artificial intelligence in caries detection: a systematic review and meta-analysis. 2025 All outcomesDental radiography (digital periapical, bitewing, panoramic; OCT/micro-CT in one study); studies 2017-2023 across China, South Korea, Brazil, Turkey, France, USA, UK, Taiwan, Germany, India, Netherlands · Digital panoramic X-rays (Fig 4b)
~70
1 study
GRADE not reported
Show 1 more finding
Accuracy of artificial intelligence in caries detection: a systematic review and meta-analysis. 2025 All outcomesDental radiography (digital periapical, bitewing, panoramic; OCT/micro-CT in one study); studies 2017-2023 across China, South Korea, Brazil, Turkey, France, USA, UK, Taiwan, Germany, India, Netherlands · Overall (all X-ray types pooled)
OR 2.72 [14.65, 14.65]
14 studies
GRADE not reported
Sensitivity · Patients undergoing dental X-ray (radiographic) imaging for caries detection
Pooling 9 studies (17,190 participants), AI sensitivity for caries detection showed no significant association (pooled OR 1.258, 95% CI 0.493-2.540) with high heterogeneity (I2=86%); ungraded.
1 finding from 1 review
Diagnostic accuracy for root caries detection · Dental radiographic/intraoral images of human teeth (X-ray and photographic)
One study reported machine-learning accuracy for root caries detection (97.1%, support vector machine); not pooled and not graded.
1 finding from 1 review
Specificity for detecting all (initial and advanced) occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the conventional (all-studies) meta-analysis, pooled specificity of visual detection for all (initial and advanced) occlusal caries was 75.3% (95% CI 70 to 80) (ungraded).
3 findings from 1 review
Specificity 75.3 [70, 80]
67 studies
GRADE not reported
Specificity 91.2 [88, 94]
12 studies
GRADE not reported
Specificity 77.0 [73, 81]
51 studies
GRADE not reported
PPV for detecting advanced occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the German private-practice setting, the positive predictive value of visual detection for advanced occlusal caries was 77.9% at an assumed prevalence of 22.0% (ungraded).
2 findings from 1 review
PPV 77.9
50 studies
GRADE not reported
PPV 30.6
55 studies
GRADE not reported
Sensitivity for detecting all (initial and advanced) occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the meta-analysis tailored to Germany, pooled sensitivity of visual detection for all (initial and advanced) occlusal caries was 49.1% (95% CI 40 to 58) (ungraded).
3 findings from 1 review
Sensitivity 49.1 [40, 58]
12 studies
GRADE not reported
Sensitivity 85.5 [81, 89]
67 studies
GRADE not reported
Sensitivity 85.8 [82, 89]
51 studies
GRADE not reported
Negative LR for detecting all (initial and advanced) occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the conventional (all-studies) meta-analysis, the negative lr of visual detection for all (initial and advanced) occlusal caries was 0.19 (95% CI 0.15 to 0.25) (ungraded).
3 findings from 1 review
Negative LR 0.19 [0.15, 0.25]
67 studies
GRADE not reported
Negative LR 0.18 [0.14, 0.24]
51 studies
GRADE not reported
Negative LR 0.56 [0.46, 0.68]
12 studies
GRADE not reported
Positive LR for detecting advanced occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the conventional (all-studies) meta-analysis, the positive lr of visual detection for advanced occlusal caries was 7.09 (95% CI 5.6 to 8.9) (ungraded).
3 findings from 1 review
Positive LR 7.09 [5.6, 8.9]
94 studies
GRADE not reported
Positive LR 5.0 [4.3, 5.8]
55 studies
GRADE not reported
Positive LR 12.5 [9.2, 16.9]
50 studies
GRADE not reported
Negative LR for detecting advanced occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the meta-analysis tailored to Germany, the negative lr of visual detection for advanced occlusal caries was 0.58 (95% CI 0.5 to 0.7) (ungraded).
3 findings from 1 review
Negative LR 0.58 [0.5, 0.7]
50 studies
GRADE not reported
Negative LR 0.39 [0.3, 0.5]
94 studies
GRADE not reported
Negative LR 0.29 [0.2, 0.4]
55 studies
GRADE not reported
NPV for detecting all (initial and advanced) occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the Egyptian public health clinic setting, the negative predictive value of visual detection for all (initial and advanced) occlusal caries was 97.4% at an assumed prevalence of 12.9% (ungraded).
2 findings from 1 review
NPV 97.4
51 studies
GRADE not reported
NPV 68.6
12 studies
GRADE not reported
Specificity for detecting advanced occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the meta-analysis tailored to Germany, pooled specificity of visual detection for advanced occlusal caries was 96.5% (95% CI 95 to 97) (ungraded).
3 findings from 1 review
Specificity 96.5 [95, 97]
50 studies
GRADE not reported
Specificity 90.9 [88, 93]
94 studies
GRADE not reported
Specificity 84.9 [82, 89]
55 studies
GRADE not reported
NPV for detecting advanced occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the German private-practice setting, the negative predictive value of visual detection for advanced occlusal caries was 85.9% at an assumed prevalence of 22.0% (ungraded).
2 findings from 1 review
NPV 85.9
50 studies
GRADE not reported
NPV 97.5
55 studies
GRADE not reported
Positive LR for detecting all (initial and advanced) occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the meta-analysis tailored to Egypt, the positive lr of visual detection for all (initial and advanced) occlusal caries was 3.73 (95% CI 3.1 to 4.4) (ungraded).
3 findings from 1 review
Positive LR 3.73 [3.1, 4.4]
51 studies
GRADE not reported
Positive LR 5.56 [3.7, 8.5]
12 studies
GRADE not reported
Positive LR 3.46 [2.8, 4.3]
67 studies
GRADE not reported
PPV for detecting all (initial and advanced) occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the German private-practice setting, the positive predictive value of visual detection for all (initial and advanced) occlusal caries was 82.0% at an assumed prevalence of 45.0% (ungraded).
2 findings from 1 review
PPV 82.0
12 studies
GRADE not reported
PPV 35.6
51 studies
GRADE not reported
Sensitivity for detecting advanced occlusal caries · Occlusal surfaces of permanent molars (demonstrated in 12-14 year-olds; underlying diagnostic accuracy studies, mostly in vitro)
In the meta-analysis tailored to Germany, pooled sensitivity of visual detection for advanced occlusal caries was 43.7% (95% CI 37 to 51) (ungraded).
3 findings from 1 review
Sensitivity 43.7 [37, 51]
50 studies
GRADE not reported
Sensitivity 64.6 [57, 71]
94 studies
GRADE not reported
Sensitivity 75.1 [70, 81]
55 studies
GRADE not reported
All occlusal carious lesions · Permanent molar surfaces in diagnostic accuracy studies (mostly in vitro; clinical)
Pooled specificity for visual detection of all occlusal lesions was 0.770 (95% CI 0.73 to 0.81).
2 findings from 1 review
Specificity 77.0 [73, 81]
51 studies
GRADE not reported
Specificity 75.3 [70, 80]
67 studies
GRADE not reported
Advanced occlusal carious lesions · Permanent molar surfaces in diagnostic accuracy studies (mostly in vitro; clinical)
Pooled specificity for visual detection of advanced occlusal lesions was 0.849 (95% CI 0.82 to 0.89).
2 findings from 1 review
Specificity 84.9 [82, 89]
55 studies
GRADE not reported
Specificity 90.9 [88, 93]
94 studies
GRADE not reported