Non-Cochrane · DTA · PM41744917
Artificial Intelligence Models for the Detection and Quantification of Orthodontically Induced Root Resorption Using Cone-Beam Computed Tomography: A Systematic Review and Meta-Analysis.
Ardila Carlos M, Pineda-Vélez Eliana, Vivares-Builes Anny M · Dentistry journal · 2026 · PubMed · doi:10.3390/dj14020079
Findings
Population Patients undergoing orthodontic treatment with roots assessed on CBCT
5 findings
| AI models (CNN or other deep learning) applied to CBCT for detection and quantification of OIRR | Manual CBCT interpretation or manual measurement/segmentation (reference standard) | Pooled sensitivity for detection of OIRR | Sensitivity 0.903 [0.818, 0.989] GRADE not reported 3 studies |
| AI models (CNN or other deep learning) applied to CBCT for detection and quantification of OIRR | Manual CBCT interpretation or manual measurement/segmentation (reference standard) | Pooled sensitivity for detection of OIRR | Sensitivity 0.903 [0.818, 0.989] Moderate Downgraded: imprecision, inconsistency 3 studies |
| AI models (CNN or other deep learning) applied to CBCT for detection and quantification of OIRR | Manual CBCT interpretation or manual measurement/segmentation (reference standard) | Pooled sensitivity for detection of OIRR | Sensitivity 0.93 [0.89, 0.96] Moderate Downgraded: imprecision, inconsistency 5 studies · 420 participants |
| AI models (CNN or other deep learning) applied to CBCT for detection and quantification of OIRR | Manual CBCT interpretation or manual measurement/segmentation (reference standard) | Agreement of AI with manual quantification of OIRR (ICC) | ICC 1.000 High 3 studies · 160 participants |
| AI models (CNN or other deep learning) applied to CBCT for detection and quantification of OIRR | Manual CBCT interpretation or manual measurement/segmentation (reference standard) | Discriminative accuracy for OIRR detection (AUC) Hybrid AI architectures (CNN combined with attention or ensemble modules) | AUC 0.955 Low Downgraded: imprecision 2 studies |
Included primary studies 7
Open in Studies| Study | Year | Also included by | Outcomes it informs here | Trial data |
|---|---|---|---|---|
| Estrella 2025 AI-AIDED VOLUMETRIC ROOT RESORPTION ASSESSMENT FOLLOWING PERSONALIZED FORCES IN ORTHODONTICS: PRELIMINARY RESULTS OF A RANDOMIZED CLINICAL… PMID:40335201 | 2025 | this review only | Discriminative accuracy for OIRR detection (AUC) | not extracted yet |
| Huang 2025 Three-dimensional partitioning and quantification of orthodontic root resorption via automatic root extraction from cone-beam computed… PMID:40055775 | 2025 | this review only | Agreement of AI with manual quantification of OIRR (ICC); PM41744917-P2-O1 | not extracted yet |
| Lin 2025 Quantitative analysis and clinical determinants of orthodontically induced root resorption using automated tooth segmentation from CBCT… PMID:40340630 | 2025 | this review only | Agreement of AI with manual quantification of OIRR (ICC); PM41744917-P2-O1 | not extracted yet |
| Pirayesh 2024 A hierarchical deep learning approach for diagnosing impacted canine-induced root resorption via cone-beam computed tomography. PMID:39180070 | 2024 | this review only | not stated | not extracted yet |
| Xu 2024 An Automatic Grading System for Orthodontically Induced External Root Resorption Based on Deep Convolutional Neural Network. PMID:38393620 | 2024 | this review only | Pooled sensitivity for detection of OIRR; Discriminative accuracy for OIRR detection (AUC) | not extracted yet |
| Xu 2025 Automatic detection of orthodontically induced external root resorption based on deep convolutional neural networks using CBCT images. PMID:40594587 | 2025 | this review only | Pooled sensitivity for detection of OIRR; Discriminative accuracy for OIRR detection (AUC) | not extracted yet |
| Zheng 2025 Automatic 3-dimensional quantification of orthodontically induced root resorption in cone-beam computed tomography images based on deep… PMID:39503671 | 2025 | this review only | Pooled sensitivity for detection of OIRR; Agreement of AI with manual quantification of OIRR (ICC); Discriminative accuracy for OIRR detection (AUC); PM41744917-P2-O1 | not extracted yet |
Review question
| Element | Planned (methods) | Found (results) |
|---|---|---|
| Population | Patients undergoing orthodontic treatment whose dental roots were assessed using CBCT | Orthodontic patients with CBCT (7 studies; sample sizes from 60 to 3000 tooth images) |
| Intervention | AI models (CNNs or other deep learning systems) for detection and quantification of OIRR | AI models: 3D U-Net (n=2), EfficientNet-B1, ensemble CNN, multi-scale CNN, custom CNN, CNN |
| Comparison | Conventional methods: manual CBCT interpretation, visual assessment by orthodontists, traditional radiographic measurements | Manual segmentation, manual measurement or manual image classification as reference standard |
| Outcomes | Accuracy, sensitivity, specificity, F1-score, AUC, ICC, Dice similarity coefficient, volumetric agreement | Sensitivity (3 studies pooled), AUC (narrative; subgroup means), ICC (3 studies pooled), Dice, accuracy, F1-score |
| Study designs | Cross-sectional, cohort, case-control studies or randomized trials using real patient CBCT data | Cross-sectional, retrospective, diagnostic performance, quantitative diagnostic, prospective multicenter and randomized trial designs (n=7) |
| Timing | Not specified | Not specified |