A scoping review of intraoperative AI-based clinical decision support systems, including those intended for use in cancer surgery, identified gaps in reporting of governance, regulatory compliance, equity, and oversight considerations, according to findings published in npj Digital Surgery.

“Rather than equating performance metrics with deployment readiness, we propose that ethical maturity and governance infrastructure are integral components of clinical validation in intraoperative care,” commented Amin Madani, MD, PhD, FRCSC, FACS, Director of the Surgical Artificial Intelligence Research Academy, University Health Network, Toronto, and colleagues. These elements “should evolve alongside technical innovation,” they emphasized.  

How the Review Was Conducted

The investigators searched five electronic databases for studies published between 2010 and July 2026 that evaluated AI-based clinical decision support systems used or intended for use during surgical or interventional procedures. Eligible studies were required to describe systems generating actionable intraoperative outputs, such as risk predictions, alerts, or decision guidance, with empirical or planned clinical evaluation in real operative settings.

Of the 3,020 records identified, five studies met the inclusion criteria: one completed single-center feasibility study, two prospective observational registries, and two randomized controlled trials. The four ongoing studies had a combined planned enrollment of 466 patients, with no published results available at the time of the review. Table 1 (see below) summarizes the included systems and their applications across interventional radiology, gynecologic oncology, and anesthesiology, ranging from image and video analysis to physiologic monitoring and therapeutic decision support.

The investigators assessed clinical validation approaches, deployment maturity, and diversity of testing environments, as well as gaps in ethical governance and reporting. They also developed a conceptual Ethics-Implementation Scorecard encompassing seven domains—safety and risk management, autonomy and human oversight, transparency and explainability, regulatory compliance and auditability, equity and representativeness, governance readiness, and clinical translation readiness—to provide a structured framework for assessing and reporting considerations related to the ethical implementation of intraoperative AI-based clinical decision support systems.

What the Review Found

Most studies were reported to explicitly describe human-in-the-loop operations, in which clinicians retained authority to accept, modify, or override AI-generated recommendations. None evaluated fully autonomous or closed-loop intraoperative decision-making.

Published evidence of intraoperative performance was limited to the feasibility study, which involved five analyzed patients. The investigators wrote that this highlighted “a substantial translational gap between technical innovation and prospective clinical validation.”

Reporting of governance, regulatory compliance, equity, and oversight considerations was less frequent than that of technical validation, they noted. When the proposed Ethics-Implementation Scorecard was applied to the five studies, scores ranged from 4 to 6 out of 7, reflecting the number of domains explicitly reported within a study, rather than defining a threshold for ethical acceptability or clinical deployment. Transparency and explainability and clinical translation readiness were found to be among the most consistently reported domains, whereas regulatory compliance and auditability was consistently underreported.

Clinical evaluation was geographically concentrated, with two studies originating from Italy and one each from Greece, China, and Chile; no eligible studies were identified from North America, Africa, or Oceania. The investigators noted that this limited geographic diversity may constrain generalizability and raise questions about equitable implementation.

“Together, these findings suggest that while intraoperative AI-based [clinical decision support] is beginning to emerge in clinical practice, ethical maturity and governance infrastructure have not progressed at the same pace as technical development,” the investigators concluded.

They continued, “As additional prospective evidence emerges, structured conceptual frameworks that assess ethical implementation alongside technical performance may support more transparent reporting and responsible translation of intraoperative AI-based [clinical decision support] into safe and equitable tools that support surgical care.” However, they cautioned that the Ethics-Implementation Scorecard remains hypothesis-generating and requires further refinement and external validation before it can be used for regulatory purposes.

DISCLOSURES: Dr. Madani is a consultant with Johnson and Johnson. The other study authors reported no conflicts of interest.

Table 1: Characteristics of Intraoperative AI-Based Clinical Decision Support Systems Included in the Scoping Review

DSA = digital subtraction angiography; SLN = sentinel lymph node.

ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO^®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.