This is a retrospective cohort study utilizing radiographic and computed tomography (CT) imaging data collected during routine clinical care at Schulthess Klinik Zürich. The study focuses on developing and validating artificial intelligence (AI)-based tools for the assessment of trapeziometacarpal (TMC) joint osteoarthritis (OA) and implant monitoring. The project is divided into four subprojects: (1) development of a new radiographic classification system for TMC OA, (2) automation of the classification using deep learning, (3) automated detection of implant migration, and (4) 3-dimensional (3D) reconstruction of the TMC joint from biplanar radiographs. Data will be sourced from two cohorts: patients from our clinical TMC arthroplasty registry who received the Touch implant, and patients with other wrist-related conditions who underwent radiographic imaging with a visible TMC joint. Together, these cohorts provide a broad coverage across the full spectrum of OA severity. OA-related features and implant related features will serve as the foundation for model training and validation. Also, they will be validated with CT images regarding reliability and accuracy. The resulting prototypes for automated OA staging, implant migration detection, and 3D modeling of the TMC joint are exclusively used for research purposes. Any future clinical use of these tools, including evaluation under Swissmedic (Swiss Agency for Therapeutic Products) regulations, will be part of a separate project.
Study Type
OBSERVATIONAL
Enrollment
2,500
No Intervention: Observational Cohort
Schulthess Klinik
Zurich, Canton of Zurich, Switzerland
New TMC OA classification
Create a novel classification system for TMC OA based on joint and bone features in X-rays. This system should ensure better reliability, accurately reflect joint health, and provide clinically relevant insights for treatment decisions.
Time frame: Preoperative
Automation of the new TMC OA classification
Implement a deep learning-based system to automate TMC joint OA classification in plain radiographs using convolutional neural networks for segmentation and feature extraction. This approach aims to improve consistency and reduce the need for manual annotations.
Time frame: Preoperative
Automated implant migration detection
Develop an AI-based system for detecting implant loosening and migration in postoperative radiographs, for early detection of loosening and long-term monitoring of implant stability.
Time frame: Postoperative
3D reconstruction of the TMC joint
Create a 3D reconstruction of the trapezium and the first metacarpal from biplanar radiographs. This provides advanced insights into individual trapezium geometry to improve surgical planning.
Time frame: Preoperative
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.