Information Technology Reference
In-Depth Information
Fig. 9 Comparisons of spine detection using different methods. a A scoliotic case using Method2
(a1) and the proposed method (a2). b An artifact case using Method1 (b1) and the proposed
method (b2)
shown in Fig. 9 . As shown in Fig. 9 a, the detection results of a scoliotic patient is
wrong if the spine geometry is modeled with the standard PCA-based approach. In
Fig. 9 b, the speci
by the imaging artifacts
and pathology in both C-spine and L-spine regions. Hence, the vertebrae labels are
wrong. In contrast, our method can robustly detect spine in both scenarios.
cally trained detectors are
confused
8 Conclusions
In this chapter, we introduced a method to automatically detect and label vertebrae
and inter-vertebral discs in medical images. By employing learning-based tech-
nologies, our method is generic for different imaging modalities. In order to achieve
a solution that is robust to severe diseases (e.g., scoliosis) and imaging artifacts
(e.g., metal artifacts), two novel components are designed in our system. First,we
emulate a radiologist and use a hierarchial strategy to learn detectors dedicated to
anchor (distinctive) vertebrae, bundle (non-distinctive) vertebrae and inter-vertebral
discs, respectively. At run-time, anchor vertebrae are detected concurrently to
provide redundant and distributed appearance cues robust to local imaging artifacts.
Bundle vertebrae detectors provide candidates of vertebrae with subtle appearance
differences, whose labels are mutually determined by anchor vertebrae to gain
additional robustness. Disc locations are derived from a cloud of responses from
disc detectors, which is robust to sporadic voxel-level errors. Second, owing to the
non-rigidness of spine anatomies, we employ a local articulated model to effec-
tively model the spatial relations across vertebrae and discs. The local articulated
model fuses appearance cues from different detectors in a way that is robust to
abnormal spine geometry resulting from severe diseases.
We tested our method on a large scale of CT (189) and MR (300) spine
scans. Veri
ed by experienced radiologists, our method reaches
perfect
labeling
Search WWH ::




Custom Search