Thanh Manh Nguyen

Thanh Manh Nguyen

Auto-generated profile, based on publicly available information

Thanh Manh Nguyen


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 August 2026 in The Open Dentistry Journal
Verified
Journal Article
Open access
PDF file
TNThanh Manh Nguyen
Thanh Manh Nguyen
TNThanh Chi Nguyen
Thanh Chi Nguyen
HLHong Thu Thi Le
Hong Thu Thi Le
HPHuyen Thi Phung
Huyen Thi Phung
HDHung Trieu Dang
Hung Trieu Dang
PNPhong Hoang Nguyen
Phong Hoang Nguyen
Introduction/Background: Temporomandibular Joint Osteoarthritis (TMJOA) is a common temporomandibular disorder that substantially impairs quality of life. Although cone-beam computed tomography is considered the gold standard for detecting osseous changes, its routine use is limited. Orthopantomographs are widely available but have reduced sensitivity for early TMJOA. This study aimed to evaluate the clinical utility of an artificial intelligence model for TMJOA diagnosis using panoramic images and to compare its performance with expert assessment. Materials and Methods: A total of 651 participants with clinical symptoms suggestive of TMJOA were included. An automated deep learning framework based on YOLOv11 was developed to extract regions of interest encompassing the mandibular condyle, articular fossa, and articular eminence to classify joints as normal or osteoarthritic. Model performance was assessed using five-fold cross-validation, and diagnostic metrics evaluated included accuracy, sensitivity, specificity, area under the curve, Cohen’s kappa, and McNemar’s test. Results: The AI model achieved a mean accuracy of 0.76, sensitivity of 0.63, specificity of 0.82, and an AUC of 0.72. Compared to experts, the AI demonstrated higher accuracy and sensitivity, while experts retained higher specificity (0.91 compared to 0.82). The agreement between the AI and expert diagnoses was moderate. Discussion: The AI model achieved moderate diagnostic performance, consistent with prior dental AI literature, and outperformed expert assessment in sensitivity (0.63 vs . 0.51). Conversely, the expert retained higher specificity (0.91), consistent with a conservative interpretive threshold shaped by clinical experience. This sensitivity-specificity trade-off mirrors broader patterns observed in AI-versus-human diagnostic comparisons. These findings support AI as a complementary screening tool, particularly where access to CBCT or specialist expertise is limited. Conclusion: The proposed AI model improves sensitivity for TMJOA detection on panoramic radiographs and performs comparably to expert clinicians. These findings support the potential role of AI as a cost-effective screening and decision-support tool in routine dental practice, particularly where access to CBCT is limited.