GumAgent: Towards an Accessible Gum Disease Detection Tool Leveraging Vision Language Model

Gum disease is a prevalent dental condition, affecting over half of the global population and posing significant public health challenges and economic burdens. In this study, we present GumAgent, an accessible gum disease detection agent tool that employs a comprehensive framework leveraging a Vision Language Model (VLM) for validating images before applying the application of a deep semantic segmentation model for gum disease detection. GumAgent effectively identifies and filters out non-relevant images, such as those featuring dentures or non-oral content, ensuring that only suitable intraoral photographs are processed. We explore various VLM approaches, including textual response feedback, to enhance the VLM's accuracy and responsiveness. Subsequently, the gum area checking tool is used to analyze the gum area. The images with sufficient gum area are fed into gum disease detection tool for monitoring gum health. This framework aims to improve accessibility to gum disease detection, enabling patients to conveniently and effectively monitor their dental health.