Page 21 - Read Online

Collins et al. Art Int Surg. 2025;5:254-69 https://dx.doi.org/10.20517/ais.2024.92 Page 266

[69]
recognize nerves below the surface using data from a hyperspectral camera , and more accurately
compared to a standard color camera used in laparoscopy. This has the potential to improve the safety of
inguinal hernia repair. However, this camera is a specialized device that is more expensive and bulkier than
a standard camera used in MIS, and it does not provide high frame rates. So, on the one hand, aleatory
uncertainty can be reduced using a specialized camera, but on the other hand, one must balance this with
any associated impracticalities or costs.

While various AI prediction models have been proposed in the literature, the next phase, before clinical
deployment, should also involve developing models that can convey accurate assessments of their
prediction uncertainty using model calibration . Calibration is required to ensure high uncertainty is
[70]
communicated when a prediction is likely to be incorrect.

In this sense, the AI models serve as a tool to assist surgeons, rather than as a replacement for human
expertise, who will need to assimilate the predictions and uncertainty information in their decision-making
processes. AI models will continue to improve, yet adequate training of surgeons to use them as tools, with a
transparent understanding of their inherent limitations, will be fundamental to success. This will require a
collaborative effort on both sides. AI models will be required to provide results and justification for their
results - known as “explainable AI” . AI models should also convey the sources of uncertainty.
[71]

On the other hand, surgeons will need additional training in data science and ML so they may understand
how to interpret AI-generated outputs and safely integrate AI into their decision-making processes. This is
not a simple feat. In radiology, recent studies show that, despite training, incorrect predictions can
negatively influence a clinical decision due to over-reliance on the technology. For instance, radiologists
were more likely to miss pathological findings in mammograms when an AI system also missed them .
[72]

As AI technology advances, its application in AWS is expected to increase, contributing to more
personalized and efficient surgical practices. Surgeons must remain informed about these developments and
adapt to emerging technologies to optimize patient outcomes while upholding established standards of care.

DECLARATIONS
Authors’ contributions
Initial manuscript drafting: Collins T
Manuscript revisions, proofreading and finalization: Saibro G, Forgione A, Hostettler A, Marescaux J

Availability of data and materials
Not applicable.

Financial support and sponsorship
None.

Conflicts of interest
Collins T is a Junior Editorial Board member of Artificial Intelligence Surgery. Marescaux J is the Honorary
Regional Editor of Artificial Intelligence Surgery. Collins T and Marescaux J are Guest Editors of the Special
Issue, The Role of Artificial Intelligence in Abdominal Wall Surgery. Collins T and Marescaux J were not
involved in any steps of editorial processing, notably including reviewer selection, manuscript handling, and
decision making. The other authors declared that there are no conflicts of interest.