Non-Cochrane · DTA · PM42231323
Artificial intelligence for predicting the pubertal growth spurt using cephalometric and hand-wrist radiographs: a systematic review and meta-analysis.
Chaves Jordana, Ruffini Matheus, Souza Adolfo, et al. · BMC oral health · 2026 · PubMed · doi:10.1186/s12903-026-08627-6
Findings
Population Children, adolescents and young adults aged 21 years or younger
32 findings
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation | Sensitivity (HSROC) 0.76 [0.70, 0.82] GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Bivariate pooled sensitivity Orthodontic radiographic datasets, mostly internal validation | Sensitivity (bivariate) 0.86 [0.78, 0.91] GRADE not reported 20 studies · 3200 participants |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation | Specificity (HSROC) 0.88 [0.83, 0.91] GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC accuracy parameter theta Orthodontic radiographic datasets, mostly internal validation | HSROC theta 0.23 [-0.01, 0.42] GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Pooled accuracy Orthodontic radiographic datasets, mostly internal validation | Accuracy (proportion) 0.83 [0.76, 0.89] GRADE not reported 21 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Bivariate pooled specificity Orthodontic radiographic datasets, mostly internal validation | Specificity (bivariate) 0.93 [0.90, 0.95] GRADE not reported 20 studies · 5071 participants |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC threshold parameter lambda Orthodontic radiographic datasets, mostly internal validation | HSROC lambda 1.91 GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Pre-spurt stage | Sensitivity (HSROC) 0.80 [0.76, 0.83] GRADE not reported 19 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Pre-spurt stage | Specificity (HSROC) 0.85 [0.81, 0.88] GRADE not reported 19 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC accuracy parameter theta Orthodontic radiographic datasets, mostly internal validation · Pre-spurt stage | HSROC theta 0.01 [-0.14, 0.16] GRADE not reported 19 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Post-spurt stage | Sensitivity (HSROC) 0.78 [0.71, 0.85] GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Post-spurt stage | Specificity (HSROC) 0.87 [0.81, 0.92] GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC accuracy parameter theta Orthodontic radiographic datasets, mostly internal validation · Post-spurt stage | HSROC theta 0.14 [-0.16, 0.40] GRADE not reported 20 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Imaging modality: CVM on cephalometric radiographs | Sensitivity (HSROC) 0.72 [0.59, 0.81] GRADE not reported 15 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Imaging modality: CVM on cephalometric radiographs | Specificity (HSROC) 0.89 [0.83, 0.93] GRADE not reported 15 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC accuracy parameter theta Orthodontic radiographic datasets, mostly internal validation · Imaging modality: CVM on cephalometric radiographs | HSROC theta 0.34 [0.05, 0.76] GRADE not reported 15 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC accuracy parameter theta Orthodontic radiographic datasets, mostly internal validation · Imaging modality: hand-wrist radiographs | HSROC theta 0.13 GRADE not reported 6 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Imaging modality: hand-wrist radiographs | Sensitivity (HSROC) 0.78 [0.63, 0.88] GRADE not reported 6 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Imaging modality: hand-wrist radiographs | Specificity (HSROC) 0.82 [0.69, 0.88] GRADE not reported 6 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Algorithm type: deep learning (CNN, Transformer, YOLO) | Sensitivity (HSROC) 0.72 GRADE not reported 13 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Algorithm type: deep learning (CNN, Transformer, YOLO) | Specificity (HSROC) 0.88 GRADE not reported 13 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC accuracy parameter theta Orthodontic radiographic datasets, mostly internal validation · Algorithm type: deep learning (CNN, Transformer, YOLO) | HSROC theta 0.35 GRADE not reported 13 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Algorithm type: traditional machine learning (SVM, MLP, ensemble) | Specificity (HSROC) 0.84 GRADE not reported 5 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Algorithm type: traditional machine learning (SVM, MLP, ensemble) | Sensitivity (HSROC) 0.76 GRADE not reported 5 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary sensitivity Orthodontic radiographic datasets, mostly internal validation · Algorithm type: large language models | Sensitivity (HSROC) 0.57 GRADE not reported 2 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | HSROC summary specificity Orthodontic radiographic datasets, mostly internal validation · Algorithm type: large language models | Specificity (HSROC) 0.78 GRADE not reported 2 studies |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Meta-regression: log sample size Orthodontic radiographic datasets, mostly internal validation · Log-transformed sample size | QM (meta-regression test of moderators) 0.51 GRADE not reported Studies not reported |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Meta-regression: sample size Orthodontic radiographic datasets, mostly internal validation · Sample size | QM (meta-regression test of moderators) 0.95 GRADE not reported Studies not reported |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Meta-regression: publication year Orthodontic radiographic datasets, mostly internal validation · Publication year | QM (meta-regression test of moderators) 0.02 GRADE not reported Studies not reported |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Meta-regression: AI architecture Orthodontic radiographic datasets, mostly internal validation · AI architecture (accounted for about 14.2% of between-study variance) | QM (meta-regression test of moderators) 5.19 GRADE not reported Studies not reported |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Meta-regression: study design Orthodontic radiographic datasets, mostly internal validation · Study design | QM (meta-regression test of moderators) 0.30 GRADE not reported Studies not reported |
| AI models (best model per study) applied to lateral cephalometric and/or hand-wrist radiographs to identify the pubertal growth spurt (three-class pre-spurt, spurt, post-spurt scheme) | Reference staging by expert assessment (CVM, Fishman SMI, Grave and Brown, DRU or other validated standards) | Meta-regression: imaging modality Orthodontic radiographic datasets, mostly internal validation · Imaging modality | QM (meta-regression test of moderators) 1.67 GRADE not reported Studies not reported |
Included primary studies 28
Open in Studies| Study | Year | Also included by | Outcomes it informs here | Trial data |
|---|---|---|---|---|
| Gonca 2024 Personality measurements and ESP tests with cards and drawings. PMID:20242024 | 1947 | 1 other review | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | not extracted yet |
| Kim 2021a Analysis of copy number variations among diverse cattle breeds. PMID:20212021 | 2010 | 1 other review | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | not extracted yet |
| Akpinar 2025 Chat Generative Pretrained Transformer-4.0’s accuracy in assessing cervical vertebrae and hand-wrist maturation stages: A retrospective… DOI:10.1016/j.ajodo.2025.08.010 | 2025 | this review only | Pooled accuracy | not extracted yet |
| Akpinar 2025 Poliomyelitis in Iowa and the United States. PMID:20252025 | 1947 | this review only | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | not extracted yet |
| Gonca 2024 Determination of growth and developmental stages in hand–wrist radiographs DOI:10.1007/s00056-023-00510-1 | 2024 | this review only | Pooled accuracy | not extracted yet |
| Jiang 2025 Deep learning based quantitative cervical vertebral maturation analysis DOI:10.1186/s13005-025-00498-6 | 2025 | this review only | not stated | not extracted yet |
| Kanchanapiboon 2024 Classification of cervical vertebral maturation stages with machine learning models: leveraging datasets with high inter- and… DOI:10.1186/s40510-024-00535-1 | 2024 | this review only | Pooled accuracy | not extracted yet |
| Kavasoglu 2025 matched by author and year (probable) | NR | this review only | Pooled accuracy | not extracted yet |
| Khazaei 2023 Automatic determination of pubertal growth spurts based on the cervical vertebral maturation staging using deep convolutional neural… DOI:10.1016/j.ejwf.2023.02.003 | 2023 | this review only | Pooled accuracy | not extracted yet |
| Khazaei 2023 [Evaluation of pharmacodynamic target attainment with vancomycin treatment of bacteremia due to Staphylococcus aureus methicillin… PMID:20232023 | 2010 | this review only | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | not extracted yet |
| Kim 2021a Estimating Cervical Vertebral Maturation with a Lateral Cephalogram Using the Convolutional Neural Network. PMID:34830682 | 2021 | this review only | Pooled accuracy | not extracted yet |
| Kim 2021b Prediction of hand‐wrist maturation stages based on cervical vertebrae images using artificial intelligence DOI:10.1111/ocr.12514 | 2021 | this review only | Pooled accuracy | not extracted yet |
| Kim 2023 Prediction of Fishman’s skeletal maturity indicators using artificial intelligence DOI:10.1038/s41598-023-33058-6 | 2023 | this review only | Pooled accuracy | not extracted yet |
| Kok 2020a Taxonomic and functional diversity of atrazine-degrading bacterial communities enriched from agrochemical factory soil. PMID:20202020 | 2010 | this review only | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | not extracted yet |
| Kok 2020a Evaluation of the Artificial Neural Network and Naive Bayes Models Trained with Vertebra Ratios for Growth and Development Determination. PMID:33828872 | 2021 | this review only | Pooled accuracy | not extracted yet |
| Kok 2021b Determination of growth and development periods in orthodontics with artificial neural network DOI:10.1111/ocr.12443 | 2020 | this review only | Pooled accuracy | not extracted yet |
| Li 2023 The psc-CVM assessment system: A three-stage type system for CVM assessment based on deep learning DOI:10.1186/s12903-023-03266-7 | 2023 | this review only | Pooled accuracy | not extracted yet |
| Liu 2024 Automated Distal Radius and Ulna Skeletal Maturity Grading from Hand Radiographs with an Attention Multi-Task Learning Method. PMID:39728901 | 2024 | this review only | Pooled accuracy | not extracted yet |
| Mohammad-Rahimi 2022 Continuous analysis and monitoring of snores and their relationship to the apnea-hypopnea index. PMID:20222022 | 2010 | this review only | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | not extracted yet |
| Mohammad-Rahimi 2022 Deep learning for the classification of cervical maturation degree and pubertal growth spurts: A pilot study. PMID:35321950 | 2022 | this review only | Pooled accuracy | not extracted yet |
| Mohammed 2024 Convolutional Neural Network-Based Deep Learning Methods for Skeletal Growth Prediction in Dental Patients. PMID:39590742 | 2024 | this review only | Pooled accuracy | not extracted yet |
| Nogueira-Reis 2024 Determination of the pubertal growth spurt by artificial intelligence analysis of cervical vertebrae maturation in lateral cephalometric… DOI:10.1016/j.oooo.2024.02.017 | 2024 | this review only | Pooled accuracy | not extracted yet |
| Ramnarayan 2025 Development and validation of an artificial intelligence algorithm for cervical vertebral maturation staging using lateral cephalograms DOI:10.1016/j.mjafi.2025.08.012 | 2025 | this review only | Pooled accuracy | not extracted yet |
| Shoari 2024 Estimating mandibular growth stage based on cervical vertebral maturation in lateral cephalometric radiographs using artificial intelligence DOI:10.1186/s40510-024-00527-1 | 2024 | this review only | Pooled accuracy | not extracted yet |
| Tentas 2025 Deep Learning Based Evaluation of Skeletal Maturation: A Comparative Analysis of Five Hand‐Wrist Methods DOI:10.1111/ocr.70008 | 2025 | this review only | Pooled accuracy | not extracted yet |
| Yildirim 2025 Can AI-Based ChatGPT Models Accurately Analyze Hand-Wrist Radiographs? A Comparative Study. PMID:40564836 | 2025 | this review only | Pooled accuracy | not extracted yet |
| Zhang 2024 CDSNet: An automated method for assessing growth stages from various anatomical regions in lateral cephalograms based on deep learning DOI:10.1016/j.ejwf.2024.09.007 | 2025 | this review only | Pooled accuracy | not extracted yet |
| Zhang 2024 [The foundation of humanity]. PMID:20242025 | 1946 | this review only | Pooled accuracy; HSROC summary sensitivity; HSROC summary specificity; Bivariate pooled sensitivity; Bivariate pooled specificity | 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). Only the non-Cochrane dental reviews have study lists, so overlap with Cochrane reviews is not counted yet.
Review question
| Element | Planned (methods) | Found (results) |
|---|---|---|
| Population | Children, adolescents and young adults in the growth phase (<=21 years) | Individuals aged 4-21 years (n=21 studies) |
| Intervention | AI models applied to hand-wrist and/or lateral cephalometric radiographs | Deep learning, traditional machine learning, LLM-based and multimodal AI on cephalometric and hand-wrist radiographs (n=21) |
| Comparison | Traditional skeletal maturation assessment (Fishman SMI, CVM or other validated standards) | CVM, Fishman SMI, Grave and Brown, DRU and other expert staging as reference standard |
| Outcomes | Accuracy, sensitivity, specificity, AUC, F1-score for identifying the pubertal growth spurt | Pooled accuracy (21 studies); HSROC and bivariate sensitivity and specificity (20 studies) |
| Study designs | Original diagnostic accuracy studies (randomized and nonrandomized) | Retrospective and cross-sectional studies (21) |
| Timing | Not specified | Not specified |