Transactions on Machine Learning and Data Mining (ISSN: 1865-6781)
Volume 10 - Number 1 - July 2017 - Pages 25-39
Mining Motion Data of Scoliotic Spine in the Coronal Plane to Predict the Spine in Lateral Bending Positions
Athena Jalalian¹, Francis E. H. Tay¹, and Gabriel Li²
¹Department of Mechanical Engineering, National University of Singapore,
²Department of Orthopedic Surgery, National University of Singapore,
Abstract
Reducing the number of X-rays required for scoliosis surgery planning is essen-tial for mitigation of radiation risks. In this regard, we aimed to study prediction of scoliotic spine curvatures in lateral bending positions by using motion data ac-quired from sequence of the spine movement from the erect to bending positions. We utilized our patient-specific multi-body kinematic model that is a chain of mi-cro-scale motion segments (MMSs) to reconstruct the curvatures and acquire the spine motion data. Hough transform was adopted to analyze the motion data. It was found that there is an excellent linear relationship (R2 = 0.93 ± 0.09) between the motion data of MMSs placed between each two successive vertebrae. Using the linear relationships, we could make good estimates of the spine curvatures (Root-mean-square-error (RMSE) = 0.0207 mm) and the key parameters for sco-liosis surgery planning; RMSEs of curvature angle, spinal mobility, and spinal flexibility were 0.0123°, 0.0089°, and 0.0002 respectively. This study showed that scoliotic spine curvatures in the bending positions and the key parameters for surgery planning can be predicted by using X-ray of the erect spine. Such an im-portant insight can lead to reduction of the number of X-rays in scoliosis standard care to mitigate the radiation risks, which is one of the surgeons’ main priorities.
Keywords:Erect Spine, Lateral Bending Positions, Multi-body Kinematic Model, Prediction, Scoliosis, Spine Curvature, Spine Motion Data.
Download Paper
(379 KB)