Causal dynamic decision-making for robotic systems in non-Markovian high-difficulty surgery

Area of Science:

• Robotics and Artificial Intelligence
• Medical Informatics
• Surgical Process Modeling

Background:

• Traditional Markov assumption-based models fail to capture time-varying effects of intraoperative anomalies.
• This limitation hinders their application in complex procedures like neurosurgery and spinal interventions.

Purpose of the Study:

• To develop a causal modeling framework that overcomes non-Markovian limitations in surgical process modeling.
• To enable intelligent response and adaptive decision-making in surgical robots.

Main Methods:

• Developed a causal modeling framework utilizing Vector Autoregression (VAR) and Granger causality analysis.
• Constructed a causal chain: original gesture → abnormal event → recovery action.
• Validated the framework on a synthetic dataset of 10,000 samples.

Main Results:

• Achieved 95.60% accuracy in causal inference.
• Demonstrated stability with an F1 score of 95.77% at 10,000 samples.
• Recall (95.88%) exceeded precision (95.34%), prioritizing safety.

Conclusions:

• The framework effectively captures non-Markovian temporal correlations from abnormal events.
• It overcomes limitations of traditional surgical decision models.
• The non-procedure-specific design offers a versatile pathway for autonomous decision-making in diverse surgical robotic applications.