SIX-Net: Spatial-Context Information miX-up for Electrode Landmark Detection

Catheter ablation is a prevalent procedure for treating atrial fibrillation, primarily utilizing catheters equipped with electrodes to gather electrophysiological signals. However, the localization of catheters in fluoroscopy images presents a challenge for clinicians due to the complexity of the intervention processes. In this paper, we propose SIX-Net, a novel algorithm intending to localize landmarks of electrodes in fluoroscopy images precisely, by mixing up spatial-context information from three aspects: First, we propose a new network architecture specially designed for global-local spatial feature aggregation; Then, we mix up spatial correlations between segmentation and landmark detection, by sequential connections between the two tasks with the help of the Segment Anything Model; Finally, a weighted loss function is carefully designed considering the relative spatial-arrangement information among electrodes in the same image. Experiment results on the test set and two clinical-challenging subsets reveal that our method outperforms several state-of-the-art landmark detection methods ( \({\sim } 50\% \) improvement for RF and \( {\sim } 25\%\) improvement for CS).