Fast Adaptation of Pre-Operative Patient Specific Models to Real-Time Intra-Operative Volumetric Data Streams.

Image-guided catheter ablation therapy is becoming an increasingly popular treatment option for atrial fibrillation. Successful treatment relies on accurate guidance of the treatment catheter. Integration of high-resolution, pre-operative data with electrophysiology data and positional data from tracked catheters improves targeting, but lacks the means to monitor changes in the atrial wall. Intra-operative ultrasound provides a method for imaging the atrial wall, but the real-time, dynamic nature of the data makes it difficult to seamlessly integrate with the static pre-operative patient-specific model. In this work, we propose a technique which uses a self-organizing map (SOM) for dynamically adapting a pre-operative model to surface patch data. The surface patch would be derived from a segmentation of the anatomy in a real-time, intra-operative ultrasound data stream. The method is demonstrated on two regular geometric shapes as well as data simulated from a real, patient computed tomography dataset.