Pruning strategies for efficient online globally consistent mosaicking in fetoscopy

Area of Science:

• Medical imaging
• Surgical navigation
• Computational geometry

Background:

• Twin-to-twin transfusion syndrome (TTTS) involves placental vascular connections causing blood flow imbalance.
• Current fetoscopic exploration for TTTS is limited by a narrow field of view, hindering localization and overview.
• Existing accurate mapping algorithms are offline, while online methods lack precision.

Purpose of the Study:

• To develop an online, accurate mosaic creation method for real-time fetoscopic exploration during TTTS surgery.
• To enhance the localization and overview of placental vascular connections.
• To improve the efficiency and accuracy of surgical navigation in TTTS treatment.

Main Methods:

• Introduced two pruning strategies to enable sequential bundle adjustment.
• Utilized an electromagnetic tracking system to prevent matching spatially inconsistent image pairs.
• Developed 'superframes'—aggregated image representations—to reduce computational complexity for global consistency.

Main Results:

• Demonstrated a superior trade-off between efficiency and accuracy using the proposed methods.
• Quantitative and qualitative results validated the approach on synthetic and phantom datasets.
• Enabled more effective exploration and localization of intertwin vascular connections.

Conclusions:

• The novel pruning strategies and superframe aggregation facilitate accurate, online bundle adjustment for TTTS surgery.
• This approach significantly improves upon the limitations of current fetoscopic visualization techniques.
• The developed methods offer a more effective and precise tool for surgeons treating twin-to-twin transfusion syndrome.