VertNet: Accurate Vertebra Localization and Identification Network from CT Images

VertNet: Accurate Vertebra Localization and Identification Network from CT Images

Zhiming Cui^1,2,5, Changjian Li³, Lei Yang², Chunfeng Lian⁴, Feng Shi⁵

Wenping Wang², Dijia Wu², Dinggang Shen²

¹ShanghaiTech University, ²The University of HongKong, ³University College London

⁴Xi’an Jiaotong University, ⁵Shanghai United Imaging Intelligence Co. Ltd.

International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2021

Fig. 1. Typical challenging cases: (a) Spine with pathological fracture; (b) Image with metal artifacts; (c) Adjacent vertebrae having similar shape appearance; (d) Image with a limited field of view.

Abstract

Accurate localization and identification of vertebrae from CT images is a fundamental step in clinical spine diagnosis and treatment. Previous methods have made various attempts in this task; however, they fail to robustly localize the vertebrae with challenging appearance or identify vertebra labels from CT images with a limited field of view. In this paper, we propose a novel two-stage framework, VertNet, for accurate and robust vertebra localization and identification from CT images. Our method first detects all vertebra centers by a weighted voting-based localization network. Then, an identification network is designed to identify the label of each detected vertebra in leveraging the synergy of global and local information. Specifically, a bidirectional relation module is designed to learn the global correlation among vertebrae along the upward and downward directions, and a continuous label map with dense annotation is employed to enhance the feature learning in local vertebra patches. Extensive experiments on a large dataset collected from real-world clinics show that our framework can accurately localize and identify vertebrae in various challenging cases and outperforms the state-of-the-art methods.

Paper [PDF]

Data [Code and Data (coming soon...)]

Citation:
Cui, Zhiming, Changjian Li, Lei Yang, Chunfeng Lian, Feng Shi, Wenping Wang, Dijia Wu, and Dinggang Shen. "VertNet: Accurate Vertebra Localization and Identification Network from CT Images." In Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part V 24, pp. 281-290. Springer International Publishing, 2021.
(bibtex)

Algorithm pipeline

Fig. 2. An overview of the proposed VertNet for vertebra localization (Sec. 2.1) and identification (Sec. 2.2) from input CT images.

Representative Results

Fig. 3. Comparison between our results (blue) and those by Deep-HMM (yellow) against the ground-truth (GT) (red). Seven typical examples are presented: metal artifacts(1,7), pathological spines(2,3), and limited field of view (4,5,6,7). The GT label is annotated if incorrect prediction occurs. (Color figure online)