3D Pseudo-Normal Knee MRI for Trochleoplasty Planning in Trochlear Dysplasia

Overview

Trochlear dysplasia (TD) causes patellar instability and pain, often requiring trochleoplasty surgery. This study introduces a novel method to generate patient-specific pseudo-healthy 3D knee MR images using wavelet diffusion models, aiding preoperative planning and potentially improving surgical outcomes.

Background

Trochlear dysplasia is a malformation of the femoral trochlea leading to lateral patellar tracking and increased dislocation risk, especially in adolescents. Untreated TD can cause chronic pain and osteoarthritis. Trochleoplasty aims to correct this deformity but is challenged by anatomical variability and risks to femoral cartilage. Current diagnostic methods lack individualized surgical planning, often resulting in open surgeries with longer recovery. The proposed approach leverages standard MR imaging and advanced inpainting techniques to create pseudo-healthy images tailored to each patient's patella anatomy.

Data Highlights

The method involves segmenting pathological regions via Otsu thresholding and morphological operations, localizing the patella, and masking the trochlear region. A wavelet diffusion model trained on healthy MR scans then inpaints the masked area to restore a normal trochlear shape. This process uses discrete wavelet transform (DWT) to decompose images into frequency bands, enabling efficient denoising and reconstruction. The approach avoids additional CT imaging, reducing radiation exposure in young patients.

Key Findings

• Trochlear dysplasia leads to abnormal sulcus angles and lateral patellar tracking, increasing dislocation risk.
• Current trochleoplasty surgeries rely on subjective image assessment and lack patient-specific preoperative plans.
• The proposed method generates pseudo-healthy MR images by masking pathological trochlear regions and using wavelet diffusion models for inpainting.
• Discrete wavelet transform enables effective decomposition and reconstruction of MR images for this task.
• The approach uses only standard MR imaging, avoiding CT-related radiation exposure.
• These pseudo-healthy images can support individualized surgical planning and may facilitate minimally invasive trochleoplasty.

Clinical Implications

This technique offers surgeons a patient-specific visualization of the ideal trochlear shape, potentially improving preoperative planning accuracy. By reducing reliance on subjective assessments and enabling tailored corrections, it may enhance surgical outcomes and reduce the need for open procedures. Additionally, avoiding CT scans minimizes radiation exposure in predominantly young patients.

Conclusion

Generating pseudo-healthy 3D knee MR images using wavelet diffusion models represents a promising advance in planning trochleoplasty for trochlear dysplasia. This approach bridges the gap between diagnosis and precise surgical correction, potentially improving patient outcomes and enabling less invasive interventions.

References

1. Barbosa et al. 2023 -- Deep Learning for Knee Landmark Detection in TD
2. Durrer et al. 2023 -- Masking and Inpainting in Medical Imaging
3. Friedrich et al. 2023 -- Wavelet Diffusion Models for Image Inpainting
4. Fang et al. 2022 -- Patient-Specific Planning in Orthognathic Surgery