Reconstruction of 3D lumbar spine models from incomplete segmentations using landmark detection

TL;DR

The paper tackles reconstructing complete 3D lumbar spine models from incomplete vertebral body segmentations by introducing an automated landmark-based affine registration that uses eight endplate landmarks per vertebra. It defines per-vertebra transforms $T(g)$ and a global matrix $R = T(t) \cdot T(s)^{-1}$ to align complete vertebra meshes with patient-specific data, followed by elastic facet-joint alignment to ensure anatomical coherence. Compared with baselines, the proposed method achieves a mean point-to-model distance of $1.95$ mm and maintains functional-spine-unit angles with MAE around $3.4^{\circ}$, while being exceptionally fast at $0.14$ s for full lumbar registration; incorporating ICP can further improve accuracy. The approach is implemented as an open-source 3D Slicer plugin and is supported by publicly released ground-truth data, enabling rapid clinical and research workflows for spine diagnostics, treatment planning, and implant design.

Abstract

Patient-specific 3D spine models serve as a foundation for spinal treatment and surgery planning as well as analysis of loading conditions in biomechanical and biomedical research. Despite advancements in imaging technologies, the reconstruction of complete 3D spine models often faces challenges due to limitations in imaging modalities such as planar X-Ray and missing certain spinal structures, such as the spinal or transverse processes, in volumetric medical images and resulting segmentations. In this study, we present a novel accurate and time-efficient method to reconstruct complete 3D lumbar spine models from incomplete 3D vertebral bodies obtained from segmented magnetic resonance images (MRI). In our method, we use an affine transformation to align artificial vertebra models with patient-specific incomplete vertebrae. The transformation matrix is derived from vertebra landmarks, which are automatically detected on the vertebra endplates. The results of our evaluation demonstrate the high accuracy of the performed registration, achieving an average point-to-model distance of 1.95 mm. Additionally, in assessing the morphological properties of the vertebrae and intervertebral characteristics, our method demonstrated a mean absolute error (MAE) of 3.4° in the angles of functional spine units (FSUs), emphasizing its effectiveness in maintaining important spinal features throughout the transformation process of individual vertebrae. Our method achieves the registration of the entire lumbar spine, spanning segments L1 to L5, in just 0.14 seconds, showcasing its time-efficiency. Clinical relevance: the fast and accurate reconstruction of spinal models from incomplete input data such as segmentations provides a foundation for many applications in spine diagnostics, treatment planning, and the development of spinal healthcare solutions.

Figures (4)

• Figure 1: Pipeline of the reconstruction method. Step 1: By determining vertebra orientations and extracting the vertebra endplates, a set of landmarks is determined for each vertebral body. Step 2: From the landmarks, the local coordinate system of the vertebra is calculated and used to determine a transformation for aligning the corresponding vertebra. Step 3: The geometry is adapted to create a coherent spine with realistic facet joint spacing. The resulting spine consists of registered complete vertebrae, that are aligned with the segments of the patient-specific spine.
• Figure 2: (a) A set of key anatomical landmarks is determined for the vertebral body. The vertebra orientation vectors are defined for three local anatomical orientations by averaging vectors between the landmarks: (b) $\vec{x}_g$ in left-right orientation, (c) $\vec{y}_g$ in posterior-anterior orientation, (d) $\vec{z}_g$ in inferior-superior orientation.
• Figure 3: The registration matrix $R$ represents a stepwise transformation from the source mesh $s$ to the global coordinate system and then to the target mesh $t$.
• Figure 4: Comparison of model surface distances for different registration methods: (a) Our method, (b) Our method + ICP, (c) ICP, (d) ICP on VB's, (e) ALPACA on VB's. The custom vertebra models are shown with the color reflecting the distances from each mesh point to the closest surface point of the registered vertebra.