Three-Dimensional Multi-Modality Registration for Orthopaedics and Cardiovascular Settings: State-of-the-Art and Clinical Applications

Abstract

The multimodal and multidomain registration of medical images have gained increasing recognition in clinical practice as a powerful tool for fusing and leveraging useful information from different imaging techniques and in different medical fields such as cardiology and orthopedics. Image registration could be a challenging process, and it strongly depends on the correct tuning of registration parameters. In this paper, the robustness and accuracy of a landmarks-based approach have been presented for five cardiac multimodal image datasets. The study is based on 3D Slicer software and it is focused on the registration of a computed tomography (CT) and 3D ultrasound time-series of post-operative mitral valve repair. The accuracy of the method, as a function of the number of landmarks used, was performed by analysing root mean square error (RMSE) and fiducial registration error (FRE) metrics. The validation of the number of landmarks resulted in an optimal number of 10 landmarks. The mean RMSE and FRE values were 5.26 ± 3.17 and 2.98 ± 1.68 mm, respectively, showing comparable performances with respect to the literature. The developed registration process was also tested on a CT orthopaedic dataset to assess the possibility of reconstructing the damaged jaw portion for a pre-operative planning setting. Overall, the proposed work shows how 3D Slicer and registration by landmarks can provide a useful environment for multimodal/unimodal registration.

Keywords: 3D Slicer; cardiovascular; landmark; multimodal registration; orthopaedic.

Figure 1

Workflow of the temporal (a) and spatial (b) registration for the cardiovascular registration.

Figure 2

Workflow of the frame reference extraction operation for the CT (a) and for US (b): first, one volume view is selected; then, one slice is extracted from the CT/US volume selected view.

Figure 3

Workflow of the US/CT frames temporal registration.

Figure 4

Workflow of the orthopaedic registration.

Figure 5

Results of the US frames selection related to one cardiac cycle. Frame selection related to the beginning (a) and to the end (b) of the cardiac cycle. The CT (c) and US (d) frame comparison operations approximately match in time, as confirmed by the MV leaflets opening in both volumes.

Figure 6

RMSE 1 of rotation (a) and translation (b) parameters plots averaged over the number of trials, as the number of pairs of landmarks varies. FRE (c) and RMSE2 (d) plots for CB3 and CB5 of User1, averaged over the trials, as the number of pairs of landmarks varies.

Figure 7

Inter-operators comparison of the FRE (a,c) and RMSE2 (b,d) plots for patients CB3 (a,b) and CB5 (c,d), averaged over the trials, as the number of pairs of landmarks varies.

Figure 8

Box plots o f the FRE (a) and RMSE2 (b) values computed on 10 landmarks over 10 trials and related to the two users.

Figure 9

The CT/US registration results for patients CB1 (a) and CB2 (b) in axial, coronal, and sagittal view. The upper row shows the overlapping of CT and MV segmentation from the US for each plane. The bottom one shows the respective US planes.

Figure 9

The CT/US registration results for patients CB1 (a) and CB2 (b) in axial, coronal, and sagittal view. The upper row shows the overlapping of CT and MV segmentation from the US for each plane. The bottom one shows the respective US planes.

Figure 10

Three-dimensional models of the tumour mass and the jaw (a) and the detail of the healthy (b) and abnormal (c) side of the jaw.

Figure 11

Mirroring pipeline from the original slice dataset (a) generates the mirrored slice dataset (b). The following registration procedure solves the matching gap visible in the overlapping of the pre-registered mirrored and original volume (c), as reported in the overlapping of the post-registered and original volume (d).

Figure 12

Three-dimensional model of the original dataset (a), overlapping of the original and mirrored dataset (b), and overlapping of the original and registered mirrored dataset (c).