Quasi-automatic 3D reconstruction of the full spine from low-dose biplanar X-rays based on statistical inferences and image analysis

Abstract

To design a quasi-automated three-dimensional reconstruction method of the spine from biplanar X-rays as the daily used method in clinical routine is based on manual adjustments of a trained operator and the reconstruction time is more than 10 minutes per patient. Methods: The proposed method of 3D reconstruction of the spine (C3-L5) relies first on a new manual input strategy designed to fit clinicians' skills. Then, a parametric model of the spine is computed using statistical inferences, image analysis techniques and fast manual rigid registration. Results: An agreement study with the clinically used method on a cohort of 57 adolescent scoliotic subjects has shown that both methods have similar performance on vertebral body position and axial rotation (null bias in both cases and standard deviation of signed differences of 1mm and 3.5° around respectively). In average, the solution could be computed in less than 5 minutes of operator time, even for severe scoliosis. Conclusions: The proposed method allows fast and accurate 3D reconstruction of the spine for wide clinical applications and represents a significant step toward full automatization of 3D reconstruction of the spine. Moreover, it is to the best of our knowledge the first method including also the cervical spine.

Quasi-automatic early detection of progressive idiopathic scoliosis from biplanar radiography: a preliminary validation

“…The initial solution was automatically calculated through statistical inference [20,21] and image analysis [22,23]. The 3D reconstructed vertebrae thus obtained were retro-projected ( Figure 2) on the frontal radiography.…”

Comparison of 3D and 2D characterization of spinal geometry from biplanar X-rays: a large cohort study

“…This new method of quasi-automatic reconstruction and measurement, which was based on statistical inferences, image processing and machine learning, greatly reduce the time-consuming work of manual measurement on radiographs or adjusting each vertebra manually via elastic deformation for orthopedic surgeon and scientists (10). It took less than 5 minutes on average to generate a 3D reconstruction of pelvis and spine; this reconstruction allows the automatic computation of parameters on coronal, sagittal and axial plane.…”

“…A quasi-automatic 3D reconstruction and measurement of the spine was carried out in four steps using previously proposed method (10). Briefly, the operator started by identifying the acetabula, sacral endplate and the spinal midline through the center of all vertebral bodies, from the tip of the odontoid process to L5, on the frontal and lateral radiographs.…”

Automated Spinal Midline Delineation on Biplanar X-Rays Using Mask R-CNN

“…Recently, research works have aimed to automated 3D spine reconstruction [8,9,10,11]. In general, the first step is to initialize a simplified statistical shape model [6] thanks to several landmarks such as the spinal midline, passing through endplates or vertebral body centers.…”

“…In general, the first step is to initialize a simplified statistical shape model [6] thanks to several landmarks such as the spinal midline, passing through endplates or vertebral body centers. In [8], they use manual inputs while in [9] they rely on a fully convolutional neural network (CNN) combined with an additional differentiable spatial to numerical (DSNT) layer to predict the locations of landmarks. In [11], a coarse localization based on the image intensity distribution of columns and rows is used, followed by a detailed statistical shape model (SSM) for landmarks position prediction and CNN patch-based regression models to correct the landmarks.…”