Current validation practice undermines surgical AI development
Abstract
Surgical data science (SDS) is rapidly advancing, yet clinical adoption of artificial intelligence (AI) in surgery remains limited, with inadequate validation emerging as an important contributing factor. In fact, existing validation practices often neglect the temporal and hierarchical structure of intraoperative videos, producing misleading, unstable, or clinically irrelevant results. In a pioneering, consensus-driven effort, we introduce a comprehensive catalog of validation pitfalls in AI-based surgical video analysis that was derived from a multi-stage Delphi process with 92 international experts. The collected pitfalls span three categories: (1) data (e.g., incomplete annotation, spurious correlations), (2) metric selection and configuration (e.g., neglect of temporal stability, mismatch with clinical needs), and (3) aggregation and reporting (e.g., clinically uninformative aggregation, failure to account for frame dependencies in hierarchical data structures). A systematic review of surgical AI papers reveals that these pitfalls are widespread in current practice, with the majority of studies failing to account for temporal dynamics or hierarchical data structure, or relying on clinically uninformative metrics. Experiments on real surgical video datasets provide empirical evidence that ignoring temporal and hierarchical data structures can substantially understate uncertainty, obscure critical failure modes, and even alter algorithm rankings. To address these shortcomings, we provide a catalogue of best practices compiled in a multi-stage Delphi process. Together, this work provides an evidence-based framework to inform more rigorous validation of surgical video analysis algorithms and to guide future efforts in benchmarking, reporting, regulatory review, and clinical translation.
Cite
@article{arxiv.2511.03769,
title = {Current validation practice undermines surgical AI development},
author = {Annika Reinke and Ziying O. Li and Minu D. Tizabi and Pascaline André and Marcel Knopp and Mika M. Rother and Ines P. Machado and Maria S. Altieri and Deepak Alapatt and Sophia Bano and Sebastian Bodenstedt and Oliver Burgert and Elvis C. S. Chen and Justin W. Collins and Olivier Colliot and Evangelia Christodoulou and Tobias Czempiel and Adrito Das and Reuben Docea and Daniel Donoho and Qi Dou and Jennifer Eckhoff and Sandy Engelhardt and Gabor Fichtinger and Philipp Fuernstahl and Pablo García Kilroy and Stamatia Giannarou and Stephen Gilbert and Ines Gockel and Patrick Godau and Jan Gödeke and Teodor P. Grantcharov and Tamas Haidegger and Alexander Hann and Makoto Hashizume and Charles Heitz and Rebecca Hisey and Hanna Hoffmann and Arnaud Huaulmé and Paul F. Jäger and Pierre Jannin and Anthony Jarc and Rohit Jena and Yueming Jin and Leo Joskowicz and Luc Joyeux and Max Kirchner and Axel Krieger and Gernot Kronreif and Kyle Lam and Shlomi Laufer and Joël L. Lavanchy and Gyusung I. Lee and Robert Lim and Peng Liu and Hani J. Marcus and Pietro Mascagni and Ozanan R. Meireles and Beat P. Mueller and Lars Mündermann and Hirenkumar Nakawala and Nassir Navab and Abdourahmane Ndong and Juliane Neumann and Felix Nickel and Marco Nolden and Chinedu Nwoye and Namkee Oh and Nicolas Padoy and Thomas Pausch and Micha Pfeiffer and Tim Rädsch and Hongliang Ren and Nicola Rieke and Dominik Rivoir and Duygu Sarikaya and Samuel Schmidgall and Matthias Seibold and Silvia Seidlitz and Alexander Seitel and Lalith Sharan and Jeffrey H. Siewerdsen and Vinkle Srivastav and Raphael Sznitman and Russell Taylor and Thuy N. Tran and Matthias Unberath and Fons van der Sommen and Martin Wagner and Amine Yamlahi and Shaohua K. Zhou and Aneeq Zia and Amin Madani and Danail Stoyanov and Stefanie Speidel and Daniel A. Hashimoto and Fiona R. Kolbinger and Lena Maier-Hein},
journal= {arXiv preprint arXiv:2511.03769},
year = {2026}
}
Comments
Under review in Nature BME