Anatomy-Aware Inference of the 3D Standing Spine Posture from 2D Radiographs

Abstract

An important factor for the development of spinal degeneration, pain and the outcome of spinal surgery is known to be the balance of the spine. It must be analyzed in an upright, standing position to ensure physiological loading conditions and visualize load-dependent deformations. Despite the complex 3D shape of the spine, this analysis is currently performed using 2D radiographs, as all frequently used 3D imaging techniques require the patient to be scanned in a prone position. To overcome this limitation, we propose a deep neural network to reconstruct the 3D spinal pose in an upright standing position, loaded naturally. Specifically, we propose a novel neural network architecture, which takes orthogonal 2D radiographs and infers the spine’s 3D posture using vertebral shape priors. In this work, we define vertebral shape priors using an atlas and a spine shape prior, incorporating both into our proposed network architecture. We validate our architecture on digitally reconstructed radiographs, achieving a 3D reconstruction Dice of 0.95, indicating an almost perfect 2D-to-3D domain translation. Validating the reconstruction accuracy of a 3D standing spine on real data is infeasible due to the lack of a valid ground truth. Hence, we design a novel experiment for this purpose, using an orientation invariant distance metric, to evaluate our model’s ability to synthesize full-3D, upright, and patient-specific spine models. We compare the synthesized spine shapes from clinical upright standing radiographs to the same patient’s 3D spinal posture in the prone position from CT.

New method to apply the lumbar lordosis of standing radiographs to supine CT-based virtual 3D lumbar spine models

“…Although there are examples of both manual 20 and automatic 10 , 21 , 24 registration-based methods, a common limitation of those is that they employ in-house scripts and other software that are not publicly available and require a great amount of technical knowledge to use. Our method utilises commercially available software which allows the accurate 3D-2D registration of vertebrae.…”

“…This could be integrated in a third-party software solution, especially Mimics, due to its capability for workflow automation by scripting (Python) [36][37][38] . Additionally, the proposed method can be improved upon by implementing currently published but commercially unavailable automation techniques for spine segmentation 39 and inferring the 3D geometry of the spine from standing biplanar radiographs 24 .…”

“…6 These authors contributed equally: Benjamin Hajnal and Peter Endre Eltes. * email: eltespeter@yahoo.com Several methods have been proposed both for 3D-2D registration of the spine (manual and automatic) 20,21 as well as inference-based models for the reconstruction of 3D spinal anatomy from standing biplanar radiographs [22][23][24] . Published registration and reconstruction-based methods can take into account the sagittal as well as coronal alignment of the spine and are able to accurately reproduce patient-specific vertebral geometry.…”

A Retrospective Observational Study to Evaluate Adjacent Segmental Degenerative Change with the Dynesys-Transition-Optima Instrumentation System

“…In the diagnosis and understanding of degenerative spinal conditions, a plethora of analytical tools are utilized to investigate the pathogenesis [3][4][5]. Various intervention strategies are employed to manage the signs and symptoms of musculoskeletal conditions.…”

Automatic 3D reconstruction of vertebrae from orthogonal bi-planar radiographs

“…In the field of 3D reconstruction of spines, there have been a series of methods proposed 5 – 9 . For example, Pomero et al 10 and Humbert et al 11 proposed to infer the 3D vertebrae based on transversal and longitudinal information and determine the overall spine morphological information through part of parameters of vertebrae based on priori knowledge.…”