Multimodal Contrastive Pretraining of CBCT and IOS for Enhanced Tooth Segmentation

Oral diseases remain one of the most pervasive global health issues, affecting over 3.5 billion individuals, which accounts for over 43% of the global population as reported by the World Health Organization [1]. This widespread prevalence underscores the critical importance of dentistry, not only for clinical needs but also for enhancing the overall quality of life for a large portion of the global population. In modern dental practice, digital dentistry plays a crucial role in streamlining workflows and enhancing patient outcomes. Cone-Beam Computed Tomography (CBCT) visualizes 3D anatomical structures, including tooth morphology, alveolar bone, and surrounding tissues [2], while intraoral scans (IOS) provide high-resolution images of occlusal surfaces that are crucial for treatment planning and prosthesis design [3]. However, these imaging modalities still require extensive manual and time-consuming analysis to identify and plan treatments [4]. Consequently, numerous research efforts now focus on automating key tasks such as caries detection [5-7], orthodontic treatment planning [8-10], and designing dental prostheses, including implants, crowns, and bridges [11-13].