Spinal muscles play an important role in two inter-related clinical problems in the thoracolumbar spine:​ 1) age-related progressive kyphosis and 2) proximal junctional kyphosis (PJK) following correction surgery. Although these disorders occur largely in the thoracic spine and show symptoms in weight-bearing postures, studies have not investigated thoracic muscles in postures other than supine. Moreover, almost all the image-based thoracolumbar models are developed from muscle data obtained from supine imaging, which questions its credibility. Hence the objectives of this study were to i) analyze the effect of posture on thoracic spinal muscle parameters in different postures, and ii) develop a method for translation of MRI-derived spinal and muscle data into a thoracolumbar biomechanical model.

Two regions (T4-T5 and T8-T9) of the thorax of six healthy volunteers were imaged (0.5T MROpen, Paramed, Genoa, Italy) in four postures (supine, standing, sitting, and flexion). Descriptive guidelines were developed to identify and quantify three muscles- trapezius (TZ), erector spine (ES), and transversospinalis (TS) from axial MR images. Intra- and inter-segmentation repeatability was assessed using ICC(3,1). The effect of spinal level and posture on muscle parameters (cross-sectional area (CSA) and position (radius and angle)) was evaluated using 2-way repeated measures ANOVA (p<0.05). A pipeline was developed to estimate subject-specific sagittal spinal geometry in different postures in order to compute muscle line-of-action in those postures. A correction factor for the direction of MRI scan slice and muscle line-of-action was computed and applied to muscle parameters of ES and TS.

The intra- and inter-rater repeatability of segmentation were excellent. The muscle size decreased (~40%) for TZ and increased for ES (~10%) caudally. The trapezius CSA increased (~10%) in standing compared to supine due to activation. CSA of all muscles decreased (~20%) during flexion compared to neutral postures, due to passive stiffness. Although, the differences in muscle parameters were found to be small (~15%), they play an important role in controlling the model outputs. The correction factor reduced the overall magnitudes of muscle parameters by about 20%. Overall, this study contributes to the growing database of thoracic muscle literature for clinical and biomechanical modelling application