AI Surgical System Supervision Module Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The deployment of AI-based systems in operating rooms is error-prone due to lack of robustness, leading to potential errors in instrument selection and time delays, as false decisions may not be noticed until they cause problems, requiring human intervention.
Innovation Solution
A preliminary validation and approval period with human supervision is introduced, where a supervision module interprets AI decisions, allows corrections, and optimizes the AI algorithm, enabling the system to act autonomously once validated, and also trains human staff.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI-based systems are deployed in operating rooms without extensive external training, then productivity is improved by enabling autonomous operation during surgical procedures, but reliability deteriorates due to lack of robustness and potential errors in decision-making
Solution Approach 1:
The system performs preliminary training and validation actions within the OR environment before full autonomous deployment. The AI system undergoes a validation period where it operates under human supervision, allowing it to learn from real surgical workflows and be approved for autonomous operation only after meeting performance criteria, thus ensuring reliability before full productivity deployment
Solution Approach 2:
Human staff serve as an intermediary during the validation period, supervising AI decisions and providing corrections. This intermediary layer allows the system to gradually achieve reliability through supervised learning while maintaining productivity benefits, transitioning from supervised to autonomous operation as robustness is validated
2Ease of operation
If AI-based systems operate autonomously in the OR, then ease of operation is improved by reducing human intervention requirements, but device complexity increases due to the need for validation and approval mechanisms
Solution Approach 1:
The system segments the deployment process into distinct phases: initial supervised operation, validation period with human oversight, and final autonomous operation. This segmentation allows the complex validation mechanisms to be managed in stages, reducing the perceived complexity while maintaining ease of operation through progressive autonomy
3Reliability
If AI systems undergo lengthy training outside the OR, then reliability is improved through extensive validation, but loss of time occurs due to delayed deployment in actual surgical procedures
Solution Approach 1:
The system performs preliminary validation actions directly within the OR environment rather than requiring extensive external training. The AI undergoes validation during actual surgical procedures under human supervision, allowing reliability to be established in the actual operational context without time loss from separate training phases
Solution Approach 2:
The AI system learns and validates itself through real surgical data and feedback from the OR environment. By using actual surgical workflows for validation rather than external training datasets, the system achieves reliability while minimizing time loss, as the learning process occurs concurrently with deployment
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
A system for implementing an AI-based surgical system configured to predict and execute steps of a surgical procedure, the system comprising: a prediction module configured to predict a step of a surgical procedure; an input configured to receive feedback on the prediction from a user and provide the feedback to the prediction module; an execution module configured to execute the step of the surgical procedure based on the prediction and based on the feedback when the system is operating under supervision; and a supervision module configured to validate and approve the system for unsupervised use based on one or more criteria.