Bridging Centers, Enhancing Knowledge: A Collaborative Initiative on Federated Learning for Dental Implant Classification

Dental implantology faces significant challenges, with over 4,000 implant models from various manufacturers making accurate classification and pathology detection increasingly complex for clinicians. This project, titled Bridging Centers, Enhancing Knowledge, aims to develop an innovative Federated Learning (FL) framework to address these challenges. By utilizing cutting-edge artificial intelligence (AI) and fostering international collaboration, the project aspires to create a reliable, standardized method for implant detection, classification, and peri-implant pathology identification.

Key Objectives

The primary goal is to develop a multicenter FL-based AI framework capable of:

  • Differentiating dental implants from teeth and assigning accurate nomenclature (FDI standards).
  • Automatically classifying implants by morphology and identifying at least five models per manufacturer.
  • Detecting peri-implant pathologies with high accuracy, offering actionable insights to clinicians.

This collaborative study leverages FL to allow multiple centers to train a unified AI model while maintaining strict data privacy. Unlike traditional centralized methods, FL ensures that data remains local to each participating institution, complying with medical data restrictions and ethical standards.

Global Collaboration

The project brings together leading institutions and researchers in dental and computational sciences:

  • Complutense University of Madrid (Spain): Coordinating the initiative and hosting the central server for model aggregation.
  • Polytechnic University of Madrid (Spain) – GIB: Providing AI expertise and engineering support.
  • University of Zurich (Switzerland): Contributing radiographic data and validating results.
  • Fourth Military Medical University (China): Adding diverse datasets to improve model robustness.
  • University of Pennsylvania (USA): Conducting local training and testing phases to assess model performance.