Introduction: Understanding how to classify and quantify three-dimensional (3D) spinal deformities remains an open question in adolescent idiopathic scoliosis. Recent studies have investigated into pattern classification based on explicit parameters. An emerging trend however seeks to synthesize complex data in a sub-space domain to capture the intrinsic variability of anatomical shape.
Objective: The objective of this study is to analyze populations of 3D scoliotic patients using a novel geometrically intuitive statistical tool for pathological cases.
Materials and Methods: Personalized 3D reconstructions of thoracic (T)/lumbar (L) spines from two cohorts: 68 Lenke Type-1 patients and 28 surgical patients with directvertebral-derotation (DVD) manoeuvres were analyzed with a non-linear manifold embedding algorithm in order to reduce the high-dimensionality of the data, using statistical properties of neighbouring spine models. We extract sub-groups of the data from the underlying manifold structure to understand the inherent distribution.
Results: For Lenke Type-1 patients, three clusters were detected from the low-dimensional manifold of 3D models: 1-hypo-kyphotic (T) with normal to high lordosis (L) (35 cases), 2-hyper-kyphotic (T) (6 cases) and 3-hypo-kyphotic (T) with loss of lordosis (L) (27 cases) (Figure1). In the case of surgical patients with DVD, the method was able to cluster both pre and post-operative spine geometries.
Conclusion: Quantitative evaluation illustrates that the complex space of spine variability can be modeled by a low-dimensional manifold and demonstrates for the first time the existence of a third class in three-dimensional shape differences for Lenke Type-1 patients.
Significance: The manifold representation can potentially be useful for classification of 3D spinal pathologies such as idiopathic scoliosis and serve as a tool for understanding the progression of deformities in longitudinal studies.