AI and surgery - Skynet or a great opportunity?

The editorial discusses the rapid emergence of artificial intelligence (AI) in surgical care, posing the central question of how AI, machine learning (ML), and reinforcement learning (RL) can be used in current surgical practice. It outlines basic definitions of AI, ML, and RL and frames their potential roles in improving surgical decision-making and performance. The authors highlight the novelty of widely available tools like ChatGPT and DALL-E while emphasizing the growing relevance of AI in both clinical and research contexts. The piece sets the stage for examining where AI can aid preoperative planning, intraoperative guidance, minimally invasive and robotic procedures, and remote surgery. It also introduces concerns around intellectual property and the pace of AI development.

While not a systematic review, the editorial references existing applications and initiatives relevant to AI in surgery. It notes improved preoperative planning leveraging historical cases and registries, and the integration of advanced imaging with navigation and robotic assistance. It cites the UK’s Accelerating Detection of Disease program intended to harness big data and AI for early disease detection. It also references literature on challenges of deep learning and autonomous actions in surgery (Taher et al., 2022), and mentions emerging work on tracking soft tissue deformation using online learning frameworks and performance-guided laparoscopic AI systems.

- AI can aid preoperative planning by analyzing historical cases and registry data to propose optimal surgical strategies. - Navigation enhanced by advanced imaging, already in use with minimally invasive and robotic surgery, has been associated in many studies with decreased complications and improved outcomes. - AI may enable earlier diagnosis and minimally invasive interventions for conditions such as sepsis, arthritis, and vascular disease, potentially reducing unsuccessful surgeries. - Intraoperative applications include efforts to accurately track soft tissue deformation via online learning frameworks to improve guidance and navigation in minimally invasive surgery. - Performance-guided laparoscopic AI systems can provide feedback (e.g., tissue characteristics) to surgeons; robots can be programmed by demonstration, imitating expert maneuvers. - AI-assisted telesurgery is a promising near-term target, enabling remote operations through AI-enhanced interfaces. - Cautions include risks of using online AI tools for patent searches that could inadvertently place proprietary ideas into the public domain, jeopardizing patentability, and broader societal concerns about rapid advances in large language models.

The editorial argues that AI’s strengths in pattern recognition, prediction, and real-time assistance align with key surgical needs such as planning, navigation, and minimally invasive robotics, thus addressing the question of how AI can be used in current practice. While full intraoperative autonomy remains distant, incremental AI augmentation—particularly in planning, guidance, and feedback—can improve safety and outcomes. The authors balance enthusiasm with caution, emphasizing responsible innovation, data quality, and awareness of legal and ethical pitfalls (notably patent disclosure risks and calls for prudence in advancing powerful language models). Overall, AI is framed as an imminent and valuable assistant rather than a near-term replacement for surgeons.

AI represents a substantial opportunity to enhance surgical care through improved planning, guidance, and minimally invasive and robotic assistance, with telesurgery poised as a compelling application. The authors advocate for careful, responsible integration of AI, recognizing both its promise and associated risks. Future work should prioritize robust datasets and registries, validation of AI tools in clinical settings, methods to model soft tissue behavior in real time, ethical and legal frameworks (including IP protection), and clear governance around the deployment of language models in healthcare contexts.

This is an editorial, not an empirical study; it does not present original data, a formal methodology, or a systematic literature review. Assertions about improved outcomes and reduced complications are general and not supported by specific quantitative evidence within the text. The perspectives may reflect the authors’ experiences and roles, and generalizability is limited.