Page 11 - Read Online

Collins et al. Art Int Surg. 2025;5:254-69 https://dx.doi.org/10.20517/ais.2024.92 Page 256

with AWS. The following factors contribute to the absence of clinically translated solutions:

Regulatory challenges: One of the primary obstacles preventing the certification of AI devices for AWS is
the rigorous regulatory landscape in healthcare. Regulatory agencies such as the FDA and EMA require that
AI-based medical devices undergo extensive validation and testing before they can be approved for clinical
use. The approval process for medical AI technologies is complex, often involving long timelines and
substantial evidence to demonstrate safety, efficacy, and clinical relevance. Given the complexity of AWS
and the variability in patient presentations, demonstrating that AI systems can reliably function across all
potential scenarios is a significant challenge.

Complexity of surgical data: Unlike fields like radiology, where imaging data follow more standardized
protocols, surgical data - especially video data from abdominal wall surgeries, are highly complex and
variable. This variability arises from variations in camera angles, visibility issues due to bleeding or smoke,
and differences in surgical techniques, instruments, and environments. These factors make it difficult to
develop robust AI models that can reliably interpret surgical situations in real time. Moreover, the lack of
standardization in surgical data limits the ability of AI systems to generalize across different hospitals,
teams, and surgical settings.

Risk and accountability: Surgery is a high-stakes environment where mistakes can have serious
consequences. For an AI system to be certified for clinical use in AWS, it must demonstrate a high level of
accuracy, reliability, and decision-making support without compromising patient safety. The important
question of accountability arises - if an AI system were to make an incorrect recommendation, determining
responsibility for the resulting harm is complex.

Integration into clinical workflow: The integration of AI into real-world surgical environments presents
significant logistical challenges. Most hospital systems are not optimized to incorporate AI tools seamlessly
into day-to-day practice. AI systems may require integration with existing electronic health record (EHR)
systems and operating room (OR) technologies. Additionally, surgical teams must be trained to trust and
responsibly operate AI tools, which requires a significant investment in time and resources, and
performance monitoring systems must be in place to ensure the AI system continues to perform as
intended. The process of integrating AI into a hospital’s workflow, ensuring smooth communication
between AI systems and surgical teams, and aligning these tools with hospital protocols is a substantial
barrier to clinical translation.

Our article focuses on state-of-the-art research and emerging AI technologies that are currently being
researched to assist AWS, rather than their immediate clinical implementation. While AI is not yet in
routine clinical use for hernia repair, there is promising research underway that could eventually improve
surgical outcomes.

ARTIFICIAL INTELLIGENCE BACKGROUND
Key concepts
Artificial intelligence (AI), machine learning (ML), and deep learning (DL) are interconnected fields that
have revolutionized various industries and research areas in recent years. AI encompasses the broad concept
of machines performing tasks that typically require human intelligence, such as problem-solving, language
comprehension, and decision making. ML, a subset of AI, focuses on developing algorithms that enable
machines to learn from data and improve over time without explicit programming. DL , in turn, is a
[18]
specialized and highly successful form of ML that utilizes computational models, known as artificial neural