Submillimeter-Accurate 3D Lumbar Spine Reconstruction from Biplanar X-Ray Images: Incorporating a Multi-Task Network and Landmark-Weighted Loss

2025-03-18

AI Summary

Current biplanar X-ray-based 3D spinal reconstruction methods suffer from insufficient accuracy (typically >1.4 mm), failing to meet clinical requirements for weight-bearing assessment. This paper proposes a fully automatic, submillimeter (0.80 mm) lumbar spine 3D reconstruction method. First, a multi-task convolutional neural network simultaneously detects vertebral anatomical landmarks and performs lumbar segmentation. Then, a deformable statistical spine model is constructed and refined via landmark-weighted 2D–3D registration to enhance geometric consistency. To our knowledge, this is the first work integrating multi-task learning with a landmark-weighted loss function into a biplanar reconstruction framework—significantly improving anatomical robustness and enabling millisecond-level reconstruction. Validated against CT-derived ground truth, the method achieves a mean reconstruction error of 0.80 mm, representing over 40% improvement over state-of-the-art approaches and fulfilling clinical applicability criteria.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Three-dimensional reconstruction of the spine under weight-bearing conditions from biplanar X-ray images is of great importance for the clinical assessment of spinal diseases. However, the current fully automated reconstruction methods only achieve millimeter-level accuracy, making it difficult to meet clinical standards. This study developed and validated a fully automated method for high-accuracy 3D reconstruction of the lumbar spine from biplanar X-ray images. The method involves lumbar decomposition and landmark detection from the raw X-ray images, followed by a deformable model and landmark-weighted 2D-3D registration approach. The reconstruction accuracy was validated by the gold standard obtained through the registration of CT-segmented vertebral models with the biplanar X-ray images. The proposed method achieved a 3D reconstruction accuracy of 0.80mm, representing a significant improvement over the mainstream approaches. This study will contribute to the clinical diagnosis of lumbar in weight-bearing positions.

Problem

• Achieves high-accuracy 3D lumbar spine reconstruction.
• Improves automated methods for clinical spinal assessment.
• Validates accuracy using CT-segmented vertebral models.

Innovation

• Automated 3D lumbar spine reconstruction
• Deformable model and landmark detection
• Submillimeter accuracy with biplanar X-ray