Evaluating a cost-effective, web-based AI platform for lateral cephalometric analysis: acomparative in-silico study.
Authors
Affiliations (4)
Affiliations (4)
- Master's candidate, Department of Oral Medicine, Periodontology, Oral Diagnosis and Oral Radiology, Faculty of dentistry, Alexandria University, Alexandria, Egypt.
- Department of Orthodontics, Faculty of Dentistry, Alexandria University, Alexandria, Egypt.
- Department of Oral Medicine, Periodontology, Oral Diagnosis and Oral Radiology, Faculty of Dentistry, Alexandria University, Alexandria, Egypt.
- Department of Oral Medicine, Periodontology, Oral Diagnosis and Oral Radiology, Faculty of Dentistry, Alexandria University, Alexandria, Egypt. [email protected].
Abstract
Despite the current popularity of artificial intelligence (AI)- assisted cephalometric analysis, its cost precludes its widespread use in low-income countries. Free alternatives are available but mostly lack rigorous validation. Thus, this study aimed to assess the precision of AI-assisted tracing using a freely available web-based platform (WebCeph) (Artificial Intelligence Orthodontic & Orthognathic Cloud Platform, South Korea) by comparing it with the gold standard (manual tracing on OnyxCeph software) (Image Instruments, Chemnitz, Germany). It was also compared with commercially available AI-assisted tracing software (Carestream) (Carestream Dental LLC, Atlanta, GA, USA). Lateral cephalometric radiographs of 57 randomly selected patients were included in this study. The radiographs were obtained from the archives of the Radiology unit, Faculty of Dentistry, Alexandria University. Three lateral cephalometric analysis methods were used. OnyxCeph3™ software was considered the gold standard (Control group), the web-based, WebCeph™ platform, was used as an online AI-based cephalometric analysis platform and set as test group 1, and the desktop-based, Carestream imaging software V8, was used for AI cephalometric analysis and set as test group 2. Skeletal and dental measurements of Steiner and Tweed analyses that were generated using the three methods were compared. Statistically significant differences were detected between the three methods in 11 out of the 18 cephalometric measurements. Notably, in Steiner analysis, the Max1-NA angle was significantly larger in WebCeph than Onyxceph and Carestream by 2.38° and 3.41°, respectively. The Mand1-NB angle was significantly larger in Onyxceph than Carestream and WebCeph by 3.07° and 2.01°, respectively. In Tweed analysis, all the measured variables were significantly different between the three methods, except the AFH. Particularly, the values of IMPA, POr-OcP, PFH, and facial height index were significantly different in Carestream than both Onyxceph and WebCeph. Fewer and smaller discrepancies were detected between WebCeph and the manual gold standard than those detected between Carestream and the manual gold standard. The freely available AI web-based platform, WebCeph, performed comparably to manual tracing, and outperformed the commercially available AI desktop-based software, Carestream, in several key parameters, making it a viable tool for resource-limited settings.