I have developed a semi-autonomous teleoperated robotic surgery system (Rahman* et al., 2021), (Gonzalez et al., 2021) that can operate with up to 5 seconds of delay with high efficiency (Figure 2). A key challenge was the scarcity of diverse training data for learning algorithms (Rahman et al., 2021). To address this, I created the DESK (Dexterous Surgical Skills) dataset (Madapana* et al., 2019), (Gonzalez* et al., 2021), which collects robotic surgical skills from various platforms. This dataset facilitates knowledge transfer across different domains, particularly where data is scarce.
My approach enables surgeons to execute procedures remotely by translating their high-level commands into robotic actions. The system breaks down complex surgeries into smaller automated segments, known as “surgemes”, improving operation efficiency even with limited data availability. It has achieved an 87% success rate in peg transfer tasks, outperforming standard teleoperation even with substantial communication delays of 5 seconds, compared to standard teleoperation systems, which can fail with delays as short as 300 millisecond. However, these systems must operate reliably in unpredictable conditions, such as disaster zones, where they may encounter novel scenarios.
References
2021
Biomedical Journal
SARTRES: A semi-autonomous robot teleoperation environment for surgery
Md Masudur Rahman*, Mythra V Balakuntala*, Glebys Gonzalez, Mridul Agarwal, Upinder Kaur, Vishnunandan LN Venkatesh, Natalia Sanchez-Tamayo, Yexiang Xue, Richard M Voyles, Vaneet Aggarwal, and Juan Wachs
Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2021