Evaluation of spinal and pelvic parameters from three-dimensional medical images

Abstract

Medical imaging of the spine and pelvis is usually performed by conventional and digital radiography, computed tomography (CT) and magnetic resonance (MR) imaging. Radiographic imaging provides two-dimensional (2D) projection images, while CT and MR imaging provide a series of 2D cross-sections that are stacked into three-dimensional (3D) images. Although radiographic imaging is still widely used due to relatively low costs and wide area of application, a paradigm shift to the use of cross-sectional imaging has occurred due to its sensitivity, specificity, availability and trends in medical practice. In the case of spine and pelvis, 3D imaging allows the measurement of morphological properties valuable for ascertaining diagnosis, determining prognosis, and guiding treatment decisions. The morphological properties of the spine and pelvis can be described by a number of parameters that represent anatomical distances or angles. The spinal curvature is described by coronal vertebral inclination (CVI) and sagittal vertebral inclination (SVI), the rotation of a vertebra around its longitudinal axis is described by axial vertebral rotation (AVR), while the sagittal alignment of the pelvis is usually described by pelvic incidence (PI). Quantification of spinal curvature, vertebral rotation and sagittal pelvic alignment is important for a better understanding of the progression of spinal and pelvic deformities. However, the human capability to quantitatively evaluate morphological parameters from medical images is limited due to our non-systematic search patterns, similar characteristics of normal and pathological conditions, and biological variability of human anatomy. Technical limitations, such as image noise, distinctive characteristics of imaging techniques and variable positioning of the patient during image acquisition, also represent a major source of variability and introduce measurement errors. This thesis concentrates on the evaluation of spinal and pelvic parameters from 3D medical images. Spinal curvature, vertebral rotation and sagittal pelvic alignment were quantitatively evaluated by existing and novel manual and computerized measurement methods, and the results were analyzed in terms of the variability introduced by methods and observers. Although manual measurements are still widely used, the most established measurement of a specific parameter is not necessarily the best choice in terms of repeatability and reliability. Computerized measurements, which apply image processing and analysis techniques, are considerably faster and less observer-dependent than manual measurements, but currently developed to a point that still requires observer interaction, usually in the form of initialization. The reproducibility and reliability are higher when parameters are evaluated from 3D than 2D images, but 3D images are more demanding in terms of availability, manipulation and interpretation. However, 3D images help clinicians in making more accurate diagnosis and planning more effective treatment strategies for spinal and pelvic deformities, while methods for computer-assisted and computerized diagnosis are constantly being developed to aid in the interpretation of the increasing amount of medical image information. Quantitative evaluation of spinal and pelvic parameters from 3D images is a current and active topic in orthopedics and medical imaging that is far from being entirely explored, therefore opportunities for further research still exist and represent challenging tasks.