Multi-Spectral MRI-Based Radiomic Features Can Classify Femoral Loosening in Patients With Total Hip Arthroplasty.

Area of Science:

• Orthopedic Surgery
• Radiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Total hip arthroplasty (THA) is a successful treatment for joint degeneration, but implant loosening is a frequent cause for revision surgery.
• Detecting early implant loosening using conventional radiography is challenging.
• Three-dimensional multi-spectral magnetic resonance imaging (3D-MSI-MRI) offers imaging capabilities near metallic implants.

Purpose of the Study:

• To investigate the efficacy of combining 3D-MSI-MRI with radiomics for detecting femoral component loosening in THA patients.
• To evaluate the performance of zonal radiomic models in identifying implant loosening compared to models using the full proximal region.

Main Methods:

• Acquired 3D-MSI-MRI data from 76 THA patients (36 symptomatic with loosening, 40 asymptomatic).
• Segmented periprosthetic trabecular bone into Gruen zones and extracted 2D radiomic features from patches.
• Applied principal component analysis for dimensionality reduction and logistic regression with cross-validation for loosening classification.

Main Results:

• Zonal radiomic models, particularly in Gruen Zones 1, 7, and 14, achieved high predictive performance (AUC > 0.79) at larger patch sizes.
• Larger patch sizes generally improved classification accuracy, though limited by anatomical constraints.
• Feature reduction retained 11-17 principal components, indicating radiomic data redundancy.

Conclusions:

• MSI-MRI-based zonal radiomics accurately detects femoral implant loosening in THA.
• This approach provides a foundation for developing radiomic decision support tools for early, quantitative detection of aseptic loosening.
• Improved early detection of loosening can potentially enhance patient outcomes.