Panoramic Landmarks: Comparing LLM-Assisted, Manual Tracing, and Self-Directed Learning in Dental Education.
Authors
Affiliations (6)
Affiliations (6)
- Department of Oral Diagnostic Sciences, Faculty of Dentistry, SEGi University, Petaling Jaya, Selangor, Malaysia.
- Department of Oral Medicine and Radiology, Saveetha Dental College and Hospitals, Saveetha Institute of Medical and Technical Sciences, Chennai, India.
- Department of Periodontics and Implantology, Faculty of Dentistry, SEGi University, Petaling Jaya, Selangor, Malaysia.
- Center of Excellence in Precision Medicine and Digital Health, Chulalongkorn University Implant and Esthetic Center, Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand; Clinic of General, Special Care and Geriatric Dentistry, Center for Dental Medicine, University of Zurich, Zurich, Switzerland. Electronic address: [email protected].
- Faculty of Dentistry, University of Puthisastra, Phnom Penh, Cambodia.
- Faculty of Dentistry, University of Puthisastra, Phnom Penh, Cambodia; Department of Dental Research Cell, Dr. D. Y. Patil Dental College & Hospital, Dr. D. Y. Patil Vidyapeeth, Pune, India.
Abstract
Accurate identification of anatomical landmarks on panoramic radiographs is a foundational yet challenging skill in dentistry. Traditional didactic teaching often requires supplementation to achieve proficiency. This study evaluates and compares the efficacy of three supplementary learning modalities: self-directed learning (SDL), traditional manual tracing (MT), and an AI-driven approach using ChatGPT. In this prospective study, 63 third-year dental students were assigned to one of three groups (n = 21 each): SDL, MT, or ChatGPT-assisted learning. Following a theoretical lecture, students were assessed using a 30-item test immediately after the lecture (baseline) and again at a 4-week follow-up. Intra- and intergroup differences were analysed using Wilcoxon signed-rank and Kruskal-Wallis tests, respectively. Intergroup analysis demonstrated that the MT group achieved significantly higher overall scores than both the SDL and ChatGPT groups (P < .05), correctly identifying the most landmarks (26/30). Within-group analysis revealed significant improvements from baseline in the MT group for 24 landmarks (P < .05 for key structures like the hard palate and hyoid bone) and in the ChatGPT group for 16 landmarks (P < .05 for the glossopharyngeal air space). The SDL group showed no significant improvement. Notably, the ChatGPT group outperformed MT in identifying four specific landmarks, including the zygomatic process and nasopharyngeal air space. For optimal learning in dental radiology, an integrated approach is recommended. MT proved most effective overall, while ChatGPT added value for specific landmarks. Combining both methods may further enhance student proficiency. Identification of landmarks is essential for accurate diagnosis and treatment planning. This study demonstrates that MT significantly enhances landmark recognition, while ChatGPT provides supplementary value. Integrating traditional and AI-assisted methods may further strengthen dental radiology education.