Performance of the Winning Algorithms of the RSNA 2022 Cervical Spine Fracture Detection Challenge
- 1 January 2024
- journal article
- research article
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 6 (1) , e230256
Abstract
"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate and report the performance of the winning algorithms of the Radiological Society of North America Cervical Spine Fracture Artificial Intelligence (AI) Challenge. Materials and Methods The competition was open to the public on Kaggle from July 28 to October 27, 2022. 3,112 CT scans with and without cervical spine fractures were assembled from multiple sites (12 institutions across 6 continents) and prepared for the competition. The test set had 1,093 scans (private test set: n= 789; mean age 53.40 ± [SD] 22.86 years; 509 males and public test set: n = 304; mean age 52.51 ± 20.73 years; 189 males) and 847 fractures. The top 8 performing algorithms were retrospectively evaluated and the area under the receiver operating characteristic curve (AUC) value, F1 score, sensitivity, and specificity were reported. Results A total of 1,108 contestants comprising 883 teams worldwide participated in the competition. The top 8 AI models showed high mean performance: AUC value of 0.96 (95% CI 0.95-0.96); F1 score of 90% (95% CI 90%- 91%); and sensitivity of 88% (95% Cl 86%-90%) and specificity of 94% (95% CI 93%-96%). Previous models have demonstrated an AUC of 0.85, F1 score of 81%, sensitivity of 76%, and specificity of 97%. Conclusion The competition successfully facilitated development of AI models that could detect and localize cervical spine fractures on CT with high performance outcomes, which appear to exceed known values of previously reported models. Further study is needed to evaluate their generalizability in a clinical environment. ©RSNA, 2024.Keywords
This publication has 12 references indexed in Scilit:
- The RSNA Cervical Spine Fracture CT DatasetRadiology: Artificial Intelligence, 2023
- Artificial Intelligence in Fracture Detection: A Systematic Review and Meta-AnalysisRadiology, 2022
- Exploring the Role of Artificial Intelligence in an Emergency and Trauma Radiology DepartmentCanadian Association of Radiologists Journal, 2020
- Why Radiology Residents Experience Burnout and How to Fix ItAcademic Radiology, 2019
- Challenges Related to Artificial Intelligence Research in Medical Imaging and the Importance of Image Analysis CompetitionsRadiology: Artificial Intelligence, 2019
- Traumatic Fractures of the Cervical Spine: Analysis of Changes in Incidence, Cause, Concurrent Injuries, and Complications Among 488,262 Patients from 2005 to 2013World Neurosurgery, 2018
- Traumatic Spinal Cord Injury in the United States, 1993-2012JAMA, 2015
- Trauma of the spine and spinal cord: imaging strategiesEuropean Spine Journal, 2009
- Evaluation of morbidity, mortality and outcome following cervical spine injuries in elderly patientsEuropean Spine Journal, 2008
- Computed Tomography Versus Plain Radiography to Screen for Cervical Spine Injury: A Meta-Analysis2005