Hostname: page-component-848d4c4894-4hhp2 Total loading time: 0 Render date: 2024-06-03T11:11:26.908Z Has data issue: false hasContentIssue false

319 Development and Validation of an Artificial Intelligence Model to Accurately Predict Spinopelvic Parameters

Published online by Cambridge University Press:  03 April 2024

Edward S Harake
Affiliation:
University of Michigan Medical School/Michigan Institute for Clinical & Health Research
Joseph R. Linzey
Affiliation:
University of Michigan Department of Neurosurgery
Cheng Jiang
Affiliation:
University of Michigan Department of Computational Medicine and Bioinformatics
Jaes C. Jones
Affiliation:
University of Michigan Department of Neurosurgery
Rushikesh Joshi
Affiliation:
University of Michigan Department of Neurosurgery
Mark Zaki
Affiliation:
University of Michigan Department of Neurosurgery
Zachary Wilseck
Affiliation:
University of Michigan Department of Radiology
Jacob Joseph
Affiliation:
University of Michigan Department of Neurosurgery
Todd Hollon
Affiliation:
University of Michigan Department of Neurosurgery
Siri Sahib S. Khalsa
Affiliation:
The Ohio State University Department of Neurosurgery
Paul Park
Affiliation:
Semmes-Murphey Clinic
Rights & Permissions [Opens in a new window]

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the ‘Save PDF’ action button.

OBJECTIVES/GOALS: The correction of spinopelvic parameters is associated with better outcomes in patients with adult spinal deformity (ASD). This study presents a novel artificial intelligence (AI) tool that automatically predicts spinopelvic parameters from spine x-rays with high accuracy and without need for any manual entry. METHODS/STUDY POPULATION: The AI model was trained/validated on 761 sagittal whole-spine x-rays to predict the following parameters: Sagittal Vertical Axis (SVA), Pelvic Tilt (PT), Pelvic Incidence (PI), Sacral Slope (SS), Lumbar Lordosis (LL), T1-Pelvic Angle (T1PA), and L1-Pelvic Angle (L1PA). A separate test set of 40 x-rays was labeled by 4 reviewers including fellowship-trained spine surgeons and a neuroradiologist. Median errors relative to the most senior reviewer were calculated to determine model accuracy on test and cropped-test (i.e. lumbosacral) images. Intraclass correlation coefficients (ICC) were used to assess inter-rater reliability RESULTS/ANTICIPATED RESULTS: The AI model exhibited the following median (IQR) parameter errors: SVA[2.1mm (8.5mm), p=0.97], PT [1.5° (1.4°), p=0.52], PI[2.3° (2.4°), p=0.27], SS[1.7° (2.2°), p=0.64], LL [2.6° (4.0°), p=0.89], T1PA [1.3° (1.1°), p=0.41], and L1PA [1.3° (1.2°), p=0.51]. The parameter errors on cropped lumbosacral images were: LL[2.9° (2.6°), p=0.80] and SS[1.9° (2.2°), p=0.78]. The AI model exhibited excellent reliability at all parameters in both whole-spine (ICC: 0.92-1.0) and lumbosacral x-rays: (ICC: 0.92-0.93). DISCUSSION/SIGNIFICANCE: Our AI model accurately predicts spinopelvic parameters with excellent reliability comparable to fellowship-trained spine surgeons and neuroradiologists. Utilization of predictive AI tools in spine-imaging can substantially aid in patient selection and surgical planning.

Type
Informatics and Data Science
Creative Commons
Creative Common License - CCCreative Common License - BYCreative Common License - NCCreative Common License - ND
This is an Open Access article, distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives licence (https://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is unaltered and is properly cited. The written permission of Cambridge University Press must be obtained for commercial re-use or in order to create a derivative work.
Copyright
© The Author(s), 2024. The Association for Clinical and Translational Science