Non-Cochrane · DTA · PM42657008
Diagnostic Accuracy of Artificial Intelligence Models for Detecting Odontogenic Keratocytes on CBCT Imaging: A Systematic Review and Meta-Analysis.
Shoorgashti Reyhaneh, Lesan Simin, Ehsani Sarah Sadat, et al. · Health science reports · 2026 · PubMed · doi:10.1002/hsr2.73086
Findings
13 findings
| Population | ||||
|---|---|---|---|---|
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Artificial intelligence models (classical machine learning and deep learning) applied to CBCT images | Histopathological diagnosis (reference standard) | Area under the SROC curve for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets | AUC 0.828 GRADE not reported 5 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Artificial intelligence models (classical machine learning and deep learning) applied to CBCT images | Histopathological diagnosis (reference standard) | Specificity for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets | Specificity 0.92 [0.81, 0.97] GRADE not reported 5 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Artificial intelligence models (classical machine learning and deep learning) applied to CBCT images | Histopathological diagnosis (reference standard) | Sensitivity for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets | Sensitivity 0.89 [0.79, 0.95] GRADE not reported 5 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Artificial intelligence models (classical machine learning and deep learning) applied to CBCT images | Histopathological diagnosis (reference standard) | Diagnostic odds ratio for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets | DOR 87.06 [28.67, 264.31] GRADE not reported 5 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Artificial intelligence models (classical machine learning and deep learning) applied to CBCT images | Histopathological diagnosis (reference standard) | Diagnostic odds ratio for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · Common-effect model | DOR 52.81 [35.49, 78.60] GRADE not reported 5 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Deep learning (CNN-based) models applied to CBCT images | Histopathological diagnosis (reference standard) | Diagnostic odds ratio for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: deep learning CNN classifiers (Chai Z 2022, Lee JH 2020); common-effect model | DOR 40.62 [24.58, 67.14] GRADE not reported 2 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Deep learning (CNN-based) models applied to CBCT images | Histopathological diagnosis (reference standard) | Specificity for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: deep learning (Liu W 2024 nnU-Net, Chai Z 2022 and Lee JH 2020 Inception v3) | Specificity 0.89 [0.65, 0.97] GRADE not reported 3 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Deep learning (CNN-based) models applied to CBCT images | Histopathological diagnosis (reference standard) | Sensitivity for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: deep learning (Liu W 2024 nnU-Net, Chai Z 2022 and Lee JH 2020 Inception v3) | Sensitivity 0.91 [0.82, 0.96] GRADE not reported 3 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Deep learning (CNN-based) models applied to CBCT images | Histopathological diagnosis (reference standard) | Diagnostic odds ratio for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: deep learning CNN classifiers (Chai Z 2022, Lee JH 2020); segmentation-based CNN analysed as a separate subgroup | DOR 47.07 [18.18, 121.84] GRADE not reported 2 studies |
| Human subjects with odontogenic keratocyst (OKC) lesions imaged with CBCT | Classical machine learning models applied to CBCT | Histopathological reference standard | Pooled diagnostic odds ratio Classical ML subgroup | DOR 129.82 [7.09, 2375.30] GRADE not reported 3 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Classical machine learning models (SVM, SDA, radiomics soft-voting ensemble) applied to CBCT images | Histopathological diagnosis (reference standard) | Specificity for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: classical machine learning (Abdolali F 2017 SVM and SDA, Sha X 2025); 3 models from 2 studies Unresolved: the two extractors disagree on this row. | Specificity 0.95 [0.81, 0.99] GRADE not reported 2 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Classical machine learning models (SVM, SDA, radiomics soft-voting ensemble) applied to CBCT images | Histopathological diagnosis (reference standard) | Diagnostic odds ratio for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: classical machine learning (Abdolali F 2017 SVM and SDA, Sha X 2025); common-effect model Unresolved: the two extractors disagree on this row. | DOR 34.64 [14.62, 82.08] GRADE not reported 2 studies |
| Patients with jaw lesions (odontogenic keratocyst vs other odontogenic cysts/tumors, e.g. ameloblastoma) imaged with CBCT | Classical machine learning models (SVM, SDA, radiomics soft-voting ensemble) applied to CBCT images | Histopathological diagnosis (reference standard) | Sensitivity for OKC detection/classification Oral and maxillofacial radiology; CBCT datasets · AI model type: classical machine learning (Abdolali F 2017 SVM and SDA, Sha X 2025); 3 models from 2 studies Unresolved: the two extractors disagree on this row. | Sensitivity 0.86 [0.59, 0.96] GRADE not reported 2 studies |
Included primary studies 13
Open in Studies| Study | Year | Also included by | Outcomes it informs here | Trial data |
|---|---|---|---|---|
| Abdolali 2017 Automated classification of maxillofacial cysts in cone beam CT images using contourlet transformation and Spherical Harmonics. PMID:28187891 | 2017 | this review only | not stated | not extracted yet |
| Bispo 2021 Computer tomographic differential diagnosis of ameloblastoma and odontogenic keratocyst: classification using a convolutional neural… PMID:33882255 | 2021 | this review only | not stated | not extracted yet |
| Chai 2022 Improved Diagnostic Accuracy of Ameloblastoma and Odontogenic Keratocyst on Cone-Beam CT by Artificial Intelligence. PMID:35155194 | 2021 | this review only | not stated | not extracted yet |
| Committeri 2024 New perspectives in the differential diagnosis of jaw lesions: Machine learning and inflammatory biomarkers. PMID:38719192 | 2024 | this review only | not stated | not extracted yet |
| Huang 2024 Odontogenic cystic lesion segmentation on cone-beam CT using an auto-adapting multi-scaled UNet. PMID:38933446 | 2024 | this review only | not stated | not extracted yet |
| Lee 2020 Diagnosis of cystic lesions using panoramic and cone beam computed tomographic images based on deep learning neural network. PMID:31677205 | 2020 | this review only | not stated | not extracted yet |
| Liu 2024 Automatic classification and segmentation of multiclass jaw lesions in cone-beam CT using deep learning. PMID:38937280 | 2024 | this review only | not stated | not extracted yet |
| Muraoka 2025 Machine learning approach using radiomics features to distinguish odontogenic cysts and tumours DOI:10.1016/j.ijom.2025.06.020 | 2026 | this review only | not stated | not extracted yet |
| Muraoka H 2025 Machine learning approach using radiomics features to distinguish odontogenic cysts and tumours. PMID:40617787 | 2026 | this review only | not stated | not extracted yet |
| Sha 2025 CBCT radiomics features combine machine learning to diagnose cystic lesions in the jaw. PMID:40111800 | 2025 | this review only | not stated | not extracted yet |
| Song 2024 Application of machine learning in the preoperative radiomic diagnosis of ameloblastoma and odontogenic keratocyst based on cone-beam CT. PMID:38627247 | 2024 | this review only | not stated | not extracted yet |
| Yeshua 2023 Deep learning for detection and 3D segmentation of maxillofacial bone lesions in cone beam CT. PMID:37191921 | 2023 | this review only | not stated | not extracted yet |
| Yilmaz 2017 Computer-aided diagnosis of periapical cyst and keratocystic odontogenic tumor on cone beam computed tomography. PMID:28688493 | 2017 | this review only | not stated | not extracted yet |
Review question
| Element | Planned (methods) | Found (results) |
|---|---|---|
| Population | Human subjects with confirmed odontogenic keratocysts evaluated using CBCT imaging | OKC cases on CBCT, 13 to 188 OKC cases per study (n=12 studies; 5 in meta-analysis) |
| Intervention | AI models (CNNs, ResNet or other deep learning architectures) for detection or classification of OKC | Classical ML (SVM, random forest, k-NN, decision tree, XGBoost, SDA) and DL (Inception v3 CNNs, nnU-Net, OCL-Net, Mask R-CNN) (n=12) |
| Comparison | Reference standard, typically histopathological confirmation or expert radiologist interpretation | Histopathological confirmation in all studies (n=12) |
| Outcomes | Sensitivity, specificity, accuracy, precision, recall, F1-score, AUC | Pooled sensitivity 89%, specificity 92%, DOR 87.06, SROC AUC 0.828 (n=5 studies, 6 models) |
| Study designs | Diagnostic accuracy studies using CBCT and AI image analysis; reviews, commentaries, editorials, case reports and panoramic or 2D-only studies excluded | Diagnostic accuracy studies published 2017 to 2025 (n=12) |
| Timing | Not specified | Not specified |