AI-Powered Dental Hygiene Tracking with Plaque Detection
Our partner set out to build an intelligent dental care tracking app capable of detecting plaque, identifying teeth, and delivering personalized hygiene feedback.
The project required the integration of computer vision, generative AI, and gamification—all wrapped in a mobile-friendly, scalable solution.
Solution
Unidatalab developed a multi-stream AI solution to automate dental hygiene analysis using user photos. It included computer vision models for plaque and tooth detection, a generative AI module for feedback, and gamified scoring logic—all deployed in a serverless AWS infrastructure.
How it works
Tooth & Plaque Detection. The system identifies each tooth and detects plaque—especially in interdental areas and along the gumline—then visualizes plaque build-up directly on the user’s photo.
Gen AI Feedback. An LLM-generated text summary provides users with actionable hygiene insights. Prompts are flexible and can be personalized based on user needs.
Gamification & Scoring. The app calculates a dental hygiene score using the Plaque Index (PI), analyzes photo angles to build a 360° model, and uses an image quality filter to ensure usable input.
Infrastructure. All processing runs on AWS Lambda in a scalable, production-ready architecture. Additional side functionality includes a toothbrush analyzer that detects brush wear and brushing behavior.
Our challenges:
Manual hygiene tracking is difficult to maintain consistently
Users often forget or lack motivation to manually log dental care routines, leading to incomplete or unreliable hygiene records.
Existing dental care apps lack accurate visual analysis
Most apps rely on self-reporting or basic photo uploads, with no intelligent analysis of plaque or tooth-level data.
Inconsistent image quality impacts tracking reliability
User-submitted photos vary in angle, lighting, and clarity, requiring robust filtering and validation mechanisms.
Project stages
We recruited 20 volunteers and collected 8,582 dental images using four photography methods. We conducted a UX survey and annotated 1,000+ images with dental specialists using the LabelMe tool.
Built modular streams: Plaque Detection, GenAI, Gamification, and Infrastructure.
Each stream focused on a key capability, from real-time feedback to hygiene scoring and photo validation.
Summary
