**5.2 The P.A.D. taxonomy—A novel scale for automation and autonomy in the OR**

Every surgical procedure is completed based on a series of perceptions, actions and decisions made by the surgeon. These three different duties are important aspects of surgery and must be included in the conversation regarding AI and its application in surgery (**Table 5**).

Perception refers to the recognition of variables in the surgical environment. Surgeons do this instinctively using their senses. Systems sense using sensors like cameras with computer vision, heat detectors, impedance measurements, etc., to convert data from a physical environment into a computational system. As an example, basic bipolar devices transfer a fixed amount of energy through the target tissue for as long as it is activated by the surgeon regardless of the state of the tissue. While using the basic bipolar it is important for the surgeon to use their own senses to see that the tissue appears to have undergone coagulation in order to stop applying energy and prevent inadvertent injury. Over-activation after the tissue has already been coagulated will create a different path of energy transfer that could damage nearby tissues. Advanced bipolar devices, in contrast, sense the tissues impedance, regulating


#### **Table 5.**

*The PAD (perception, action, decision) scale for surgical autonomy.*

*Human-Machine Collaboration in AI-Assisted Surgery: Balancing Autonomy and Expertise DOI: http://dx.doi.org/10.5772/intechopen.111556*

the amount of energy dispensed, and automatically discontinuing the activation when the tissue is coagulated.

Action refers to the maneuvers performed in order to execute a task. Surgeons perform actions depending on their perception of a specific scenario. Basic tools and systems can perform actions without having the ability to sense. Advanced systems have the ability to perform an action depending on what they sense. The advanced bipolar device, for example, acts to continue energy or stop it according to its own perception (sensing).

Decision refers to the capability of reaching a conclusion after considering different variables. Advanced systems can give real-time feedback to the surgeon during a procedure, either passively in the form of alerts, warnings and suggestions, or actively in the form of whole system halts, or action restrictions. For example, an advanced laparoscopic stapler can sense the cartridge type inserted to the device as well as the distance and physical resistance between its two jaws. When the stapler is ready to fire, if these variables exceed the stapler's ability, it makes the decision not to fire.

With this taxonomy, one can describe easily the level of AI autonomy by combining each section into a shortened form. As such, the current standard of care is at P1A1D2, because although AI is not yet commercially available, we do have tools like advanced devices that perform certain actions autonomously. Applying the scale to these commercially available devices, we can say that advanced bipolar devices are a Level 1 automated systems as they measure the impedance of a tissue to automatically decide when a cycle is completed. A procedure using this device would therefore be characterized as a P1A1D1 procedure. Smart staplers such as the Signia™ would also be a Level 1 system and a surgeon using it would also be performing a P1A1D2 procedure. As current technologies are further developed with the evolution of AI into more clinical applications, procedures at the level of P2A1D3 may in fact be in our near future.

It is important to note that according to these Levels of Autonomy in Surgery, the responsibility still always falls upon the surgeon, independently of the amount of control and relative autonomy that the system has. The natural path of the debate in the field will bring surgeons (and healthcare professionals in general) to reach a consensus on the amount of control we are willing to give up for the whether it should be ethical and legal for a surgeon to actually relinquish control and autonomy to the point where the burden of responsibility should not be placed on them.

#### **5.3 Will AI replace surgeons?**

As with any industry, the perceived threat of AI taking control and pushing away human involvement holds true in medicine. Although at the peak of the hype of AI in radiology and pathology many experts predicted that humans would soon be replaced by machines in these fields, they quickly revised their opinions, with the realization that rather than replacement, the technology had arrived to augment their field's possibilities [67]. This is true in the surgical field as well, and as part of the adoption of AI, surgeons will have to adapt training methods to include these new systems. Not as a way of replacing, but as a way of augmenting the surgeon's capabilities. As such, it is imperative that surgeons understand the capabilities and limitations of the technology, that they know how to use it and problem solve with it, with enough exposure during their training to feel comfortable adding it to their bag of armament. More importantly, as the technology advances it remains imperative that the surgeons retain the ability to perform a surgery with all the necessary tasks safely, even without the use of automated systems. This is particularly important when faced with ensuring safety of patients: Imagine the problematic hypothetical scenario of surgeon who is unable to perform a cholecystectomy due to lack of the ability to recognize the triangle of safety because they rely solely on AI. Conversely, imagine the exciting scenario where a surgeon who is trained to recognize the triangle of safety can utilize AI tools to augment its visualization in a patient with complex anatomy, bringing an added benefit to the patient and the surgeon themselves.

Fundamentally, it is possible to continue to build on the basis of the surgeon's knowledge while maintaining control and delegating specific tasks to AI in order to augment their capabilities, not replace them. As long as the human understands the capabilities and limitations of an AI system as laid out above, the loss of control is thereby mitigated.

It cannot be stressed enough that medicine is a profession of empathy. As physicians we consider more than just the patient's diagnosis in order to propose an appropriate treatment and management. Surgeons must weigh the patient's prognosis, social support system, risks involved in surgery, and patient expectations in order to propose the best treatment. Moreover, during surgery we make an immeasurable amount of decisions and subsequent actions based on the unique patient laying on our table. We cannot say that AI has the ability to consider a patient's environment, desires and expectations, nor can we say that it is machine-proof, but the potential for an AI system with the ability to make such decisions with empathy remains only a theoretical concept for now.
