Peer-reviewedCC BY 4.0Scientific Reports
Exploring clinician interpretability of AI decision support systems for surgical anatomy recognition
Abstract
Artificial intelligence (AI) decision support systems for surgery hold promise but face barriers to adoption, particularly around the interpretability of their outputs. We conducted an international cross-sectional survey of 47 neurosurgeons to evaluate perspectives on literature-derived explanatory techniques for AI-generated anatomical segmentations, using endoscopic pituitary surgery as a high-risk exemplar. Participants ranked certainty scores, certainty maps, saliency maps, scene similarity scores, and nearest-neighbour illustrations, and rated them using a modified Explanation Satisfaction Scale alongside free-text feedback. Certainty-based techniques were consistently ranked and rated highest - valued for aligning with surgical decision-making by conveying confidence (via scores) and anatomical boundaries (via maps). Saliency- and similarity-based methods were judged less clinically relevant and better suited to educational settings. Therefore, certainty-based techniques appear most acceptable to surgeons for improving the interpretability of AI-assisted anatomical navigation systems, though their impact on trust calibration and performance requires prospective evaluation across surgical domains.
Related
- Exploring Healthy Neurocognitive Ageing with Deep Learning Interpretability Methods
- National Clinician Survey of Outcomes for Trials in Surgical Wounds Healing by Secondary Intention (SWHSI).
- From De-escalation to Restraint: A Qualitative Study of Emergency Clinician Decision-Making in Patient Agitation.
- AI-Assisted Clinical Decision Support Versus Standard Teleconsultation in Adult Medicine: A Systematic Review of Diagnostic Accuracy, Referral and Testing Decisions, and Clinician Confidence.
- Measuring what matters- patient and clinician priorities for clinical decision support interactions: a modified Delphi study.
BibTeX
@article{khan2026exploring,
title = {{Exploring clinician interpretability of AI decision support systems for surgical anatomy recognition}},
author = {Danyal Zaman Khan and Tobias Joseph Adams and Anjana Wijekoon and Roxana Ramirez Herrera and Sophia Bano and Peter McCulloch and Danail Stoyanov and Matthew J Clarkson and Enrico Costanza and Ann Blandford and Hani Joseph Marcus and CARES Evaluation Group and Mohammed Mustafa Adawi and Joao Paulo Almeida and Jensen Ang and Anouk Borg and Roman Bošnjak and Alexandros Boukas and Aswin Chari and Lekhaj C Daggubati and Neil L Dorward and Samuel Emerson and Laurence Johann Glancz and Paul Gwammache and John Gerrard Hanrahan and Joachim Starup-Hansen and Lauren Harris and Mark Hughes and Abhiney Jain and Adham M Khalafallah and Taufiq Khan and Muhammad Sohaib Khan and Ramez Wadie Kirollos and Angelos G Kolias and Jacek Kunicki and Ruth Lau and Juan Carlos Fernandez-Miranda and Ramesh Nair and Nicola Newall and Amr Nimer and Fabio Massimo Oddi and Piero Andrea Oppido and Piotr Paździora and Hafiz Muhammad Irfan Razzaq and Luis Alejandro Flores Salazar and Olabisi Sanusi and Thomas Santarius and Sebastian Senger and Jonathan Shapey and Nathan A Shlobin and Nicholas Wake Macmeikan Thomas and Sebastian Miguel Toescu and Georgios Tsermoulas and Juan Francisco Villalonga and Adam Williams and Nina Yoh and Brett E Youngerman and Gabriel Zada},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-74953-y},
url = {https://www.nature.com/articles/s41598-026-74953-y},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Exploring clinician interpretability of AI decision support systems for surgical anatomy recognition
AU - Danyal Zaman Khan
AU - Tobias Joseph Adams
AU - Anjana Wijekoon
AU - Roxana Ramirez Herrera
AU - Sophia Bano
AU - Peter McCulloch
AU - Danail Stoyanov
AU - Matthew J Clarkson
AU - Enrico Costanza
AU - Ann Blandford
AU - Hani Joseph Marcus
AU - CARES Evaluation Group
AU - Mohammed Mustafa Adawi
AU - Joao Paulo Almeida
AU - Jensen Ang
AU - Anouk Borg
AU - Roman Bošnjak
AU - Alexandros Boukas
AU - Aswin Chari
AU - Lekhaj C Daggubati
AU - Neil L Dorward
AU - Samuel Emerson
AU - Laurence Johann Glancz
AU - Paul Gwammache
AU - John Gerrard Hanrahan
AU - Joachim Starup-Hansen
AU - Lauren Harris
AU - Mark Hughes
AU - Abhiney Jain
AU - Adham M Khalafallah
AU - Taufiq Khan
AU - Muhammad Sohaib Khan
AU - Ramez Wadie Kirollos
AU - Angelos G Kolias
AU - Jacek Kunicki
AU - Ruth Lau
AU - Juan Carlos Fernandez-Miranda
AU - Ramesh Nair
AU - Nicola Newall
AU - Amr Nimer
AU - Fabio Massimo Oddi
AU - Piero Andrea Oppido
AU - Piotr Paździora
AU - Hafiz Muhammad Irfan Razzaq
AU - Luis Alejandro Flores Salazar
AU - Olabisi Sanusi
AU - Thomas Santarius
AU - Sebastian Senger
AU - Jonathan Shapey
AU - Nathan A Shlobin
AU - Nicholas Wake Macmeikan Thomas
AU - Sebastian Miguel Toescu
AU - Georgios Tsermoulas
AU - Juan Francisco Villalonga
AU - Adam Williams
AU - Nina Yoh
AU - Brett E Youngerman
AU - Gabriel Zada
PY - 2026
DA - 2026/10/24
JO - Scientific Reports
DO - 10.1038/s41598-026-74953-y
UR - https://www.nature.com/articles/s41598-026-74953-y
AB - Artificial intelligence (AI) decision support systems for surgery hold promise but face barriers to adoption, particularly around the interpretability of their outputs. We conducted an international cross-sectional survey of 47 neurosurgeons to evaluate perspectives on literature-derived explanatory techniques for AI-generated anatomical segmentations, using endoscopic pituitary surgery as a high-risk exemplar. Participants ranked certainty scores, certainty maps, saliency maps, scene similarity scores, and nearest-neighbour illustrations, and rated them using a modified Explanation Satisfaction Scale alongside free-text feedback. Certainty-based techniques were consistently ranked and rated highest - valued for aligning with surgical decision-making by conveying confidence (via scores) and anatomical boundaries (via maps). Saliency- and similarity-based methods were judged less clinically relevant and better suited to educational settings. Therefore, certainty-based techniques appear most acceptable to surgeons for improving the interpretability of AI-assisted anatomical navigation systems, though their impact on trust calibration and performance requires prospective evaluation across surgical domains.
SN - 2045-2322
ER -