Probabilistic visual and electromagnetic data fusion for robust drift-free sequential mosaicking: application to...

Area of Science:

• Medical Imaging
• Surgical Navigation
• Fetal Surgery

Background:

• Twin-to-twin transfusion syndrome (TTTS) requires precise treatment via laser photocoagulation of placental vascular anastomoses.
• Locating these fine vascular connections with a fetoscope is difficult due to limited field-of-view and vessel caliber.
• Current vision-based mosaicking techniques suffer from drift, hindering accurate navigation and treatment planning.

Purpose of the Study:

• To develop a drift-free mosaicking technique for enhanced fetoscopic visualization in TTTS treatment.
• To integrate electromagnetic tracking (EMT) with visual information for globally consistent placental mosaics.
• To improve navigation and decision-making for laser photocoagulation of placental vessels.

Main Methods:

• Proposed a probabilistic framework combining electromagnetic tracker (EMT) sensor data with frame-to-frame visual information.
• Utilized plane and camera poses constrained by EMT measurements for global consistency.
• Employed local bundle adjustment with EMT integration for sequential mosaicking.
• Evaluated the approach using synthetic, phantom-based, and ex vivo datasets.

Main Results:

• The proposed EMT-integrated mosaicking approach is drift-free.
• Achieved performance comparable to state-of-the-art global alignment techniques with reduced computational cost.
• Demonstrated robustness to EMT noise and temporary loss of visual information.
• Successfully generated consistent mosaics for various datasets.

Conclusions:

• Integrating electromagnetic tracking with visual mosaicking offers a robust and efficient solution for fetoscopic navigation in TTTS treatment.
• This drift-free method enhances the ability to locate and target placental vascular anastomoses for laser photocoagulation.
• The technique shows promise for improving surgical outcomes in complex fetal interventions.