| Home > Publications database > Artificial intelligence in cancer surgery. [KI in der onkologischen Chirurgie.] |
| Journal Article | DKFZ-2026-00389 |
; ; ; ; ; ;
2026
Springer Medizin
New York
Abstract: The integration of artificial intelligence (AI) is lagging in interventional fields due to the necessity of real-time processing of multimodal data for time-critical, often irreversible, intraoperative decisions. This stands in contrast to the established use of AI in diagnostic disciplines such as radiology and pathology. The demands on temporal latency and system robustness are significantly higher than in diagnostics. Current industrial AI applications are limited to retrospective analyses and simple tasks like “phase recognition” or automated detection of the critical view of safety (CVS) in cholecystectomy. Application in surgical oncology is more complex due to tumor infiltration and scarring blurring anatomical boundaries as well as the scarcity and heterogeneity of video data. Clinical translation is often hampered by the lack of a standardized, interoperable operating room (OR) data infrastructure; the high effort associated with consistent human annotation of surgical videos; and the absence of external and prospective validation of AI models. The development of robotic systems, ranging from assistive to fully autonomous, is progressing (e.g., autonomous ex vivo cholecystectomy). However, unresolved regulatory, ethical, and liability issues impede the clinical translation of these systems. Essential measures for advancement include building a high-performance OR data infrastructure, standardization of annotation protocols by professional societies, and rigorous external validation. In conclusion, AI has the potential to fundamentally transform surgical oncology. However, realizing its clinical benefit requires strong leadership by surgeons in close collaboration with industry and engineers.
|
The record appears in these collections: |