A novel intrarenal multimodal 2D/3D registration algorithm and preliminary phantom study.

Area of Science:

• Medical Imaging
• Surgical Navigation
• Computer-Aided Surgery

Background:

• Retrograde intrarenal surgery (RIRS) is a key procedure for upper urinary tract issues.
• Image-guided navigation systems enhance surgical precision but face challenges with complex organ structures like kidneys.
• Classical intensity-based registration methods struggle with intensity variations between virtual and real images.

Purpose of the Study:

• To develop an improved image registration method for enhanced accuracy and robustness in RIRS.
• To address the limitations of traditional registration techniques in complex anatomical environments.
• To improve the reliability of image-guided navigation in minimally invasive surgery.

Main Methods:

• A novel method combining structural feature similarity with a semantic style transfer network.
• Integration of multi-view constraints to overcome spatial depth information loss.
• Validation using patient-derived models and comparison with existing algorithms.

Main Results:

• The proposed method achieved a mean target error (mTRE) of 0.971 ± 0.585 mm and 1.266 ± 0.416 mm on two patient models.
• Demonstrated superior accuracy and robustness compared to competing algorithms, especially with significant initial deviations.
• Effectively compensated for spatial depth information collapse using multi-view constraints.

Conclusions:

• The developed structural feature similarity and semantic style transfer method significantly improves image registration accuracy for RIRS.
• The approach offers enhanced robustness and precision, overcoming challenges posed by complex anatomical structures.
• This method shows potential for application in RIRS and other surgeries involving similarly structured organs.