FPGA-Based Low-Latency Semantic Feedback System for Real-Time Instrument Localization in Telesurgery.

BACKGROUND

Telesurgery is limited by visual latency, compromising master-slave synchronization and safety. This study proposes a real-time semantic feedback framework that reduces latency while maintaining millimetre-level localization accuracy.

METHODS

We developed a low-latency semantic feedback system for real-time instrument localization using a custom visual marker and FPGA-based extraction. The system was evaluated through phantom experiments, accuracy benchmarking against the NDI system, latency measurement, and task-based evaluation under delayed visual feedback conditions.

RESULTS

The system achieved a 99.2% marker extraction success rate, with a maximum mean error of 0.68 ± $\pm $ 0.72 mm and an end-to-end latency of 31.2 ± $\pm $ 1.4 ms. Task-based evaluation further showed significantly reduced task completion time under delayed visual feedback.

CONCLUSIONS

The proposed system provides robust, precise, and low-latency semantic feedback and demonstrates potential to mitigate the impact of visual latency on surgical performance, supporting its application in future telesurgical systems pending further in vivo validation.