Communication is going to be essentially very crucial. This is especially true given that many surgical procedures are time sensitive. That means an optimization of time is critical, how different components and systems interact with each other is crucial, and that said system needs to be fault tolerant. It would also implement a closed-loop type system where everything was preconfigured as a result of a learning model in regards to said optimal result.
Something crucial to understand is that the learning mechanism will likely need to implement multiple components. When dealing with a high level and degree of sensitivity such as the brain, preconfigured imaging of the brain is likely a must. It is also important to consider the lateralization of the brain, size, and hemisphere. Size variation and difference are also important to consider. This is why there is both a computer vision aspect to it, as well as preconfigured modeling in regards to degrees of freedom and spatial density.