Home facilities research publications collaboration training staff gallery shareddata sharedcode about contact
only search CIVM

Supplemental Material for:

Registration-based segmentation of murine 4D cardiac micro-CT data using symmetric normalization

Darin Clark, Alexandra Badea, Yilin Liu, G. Allan Johnson, Cristian T. Badea

Physics in Medicine and Biology 57:6125-6145, 2012. PMCID: PMC3615410

Micro-CT can play an important role in preclinical studies of cardiovascular disease because of its high spatial and temporal resolution. Quantitative analysis of 4D cardiac images requires segmentation of the cardiac chambers at each time point, an extremely time consuming process if done manually. To improve throughput this study proposes a pipeline for registration-based segmentation and functional analysis of 4D cardiac micro-CT data in the mouse. Following optimization and validation using simulations, the pipeline was applied to in vivo cardiac micro-CT data corresponding to 10 cardiac phases acquired in C57BL/6 mice (n = 5). After edge-preserving smoothing with a novel adaptation of 4D bilateral filtration, one phase within each cardiac sequence was manually segmented. Deformable registration was used to propagate these labels to all other cardiac phases for segmentation. The volumes of each cardiac chamber were calculated and used to derive stroke volume, ejection fraction, cardiac output, and cardiac index. Dice coefficients and volume accuracies were used to compare manual segmentations of two additional phases with their corresponding propagated labels. Both measures were, on average, >0.90 for the left ventricle and >0.80 for the myocardium, the right ventricle, and the right atrium, consistent with trends in inter- and intra-segmenter variability. Segmentation of the left atrium was less reliable. On average, the functional metrics of interest were underestimated by 6.76% or more due to systematic label propagation errors around atrioventricular valves; however, execution of the pipeline was 80% faster than performing analogous manual segmentation of each phase.

Supplement to Figure 2. (A) original MOBY phantom, phase 4, (B) phase 4 as reconstructed with noise, (C) the results of applying "Median Bilateral Filtration (BF) with Registration" (the 4D filtration scheme used throughout the manuscript) to (B), and (D) the difference before and after filtration (B-C). The RMSE of the noisy image (B) relative to (A) is reported below panel (B) for reference. The RMSE of (C) relative to (A) and the mean and standard deviation (SD) of (D) are as reported below their respective panels in HU. Additional rows compare the results of related filters performed in 3D (left column) and 4D (right column) to panels (C) and (D). With the exception of the 4D "Gaussian Filter", all 4D filters are seen to perform better than their 3D counterparts (lower RMSE, lower mean difference). Red arrows illustrate oversmoothing between the left ventricle (LV) and left atrium (LA) by all 4D filters except "Classic BF" and "Median BF with Registration". Over all filters, "Median BF with Registration" is seen to minimize the RMSE, motivating our choice to use it in the manuscript. "Median BF" refers to BF in which the range weights are centered on the median intensity value within the filtration kernel (Francis and de Jager 2003) instead of on the intensity value at the kernel's center ("Classic BF", Tomasi and Manduchi 1998). All kernel sizes were fixed at 3x3x3(x3) (x, y, z, time), and the range standard deviation was fixed at 79 HU for comparison purposes. The standard deviation of the Gaussian used for the "Gaussian Filter" was 0.65 voxels with the 3D kernel replicated in time for 4D filtration. The calibration bar at the top right denotes the window width and level in HU for all grayscale and difference images. See Clark et al. 2012 for a more comprehensive discussion of 3D and 4D BF of murine 4D cardiac micro-CT data.

Figure 5. Registration Based Segmentation in Real Data: Slice views of the long axis (A) and the short axis (B) of the heart as manually segmented in phase 3 and as propagated to phases 1 and 5. Red contours indicate the boundary of the myocardium. (C) 3D rendering of the labels at each phase with matching colors by anatomy. The myocardium is semi-transparent. The calibration bar at the right of row B denotes the window width and level in HU for the grayscale images before superimposing the labels.


Note: For optimal quality, please download the following videos, and view them from your local computer.

All images and PDF documents associated with this publication are available from CIVMSpace, our Web-based data portal.

Data downloaded from this site is for academic use only. If you use this data in a publication please send us a request for copyright permission and appropriate acknowledgements. Licenses can be granted for commercial use.
Contact the Center for permission.

The images, movies and PDF documents are:

Row A:

Long Axis - Phase 1 (Label Propagation)

Long Axis - Phase 3 (Manual Segmentation)

Long Axis - Phase 5 (Label Propagation)

Row B:

Short Axis - Phase 1 (Label Propagation)

Short Axis - Phase 3 (Manual Segmentation)

Short Axis - Phase 5 (Label Progagation)

Row C:

Long Axis - 4D Animation (Complete Label Progation from phase 3)

System Requirements:
CIVMSpace is designed to work on most platforms and is supported in most browsers.

VoxStation requires a working Java installation.


  • All work was performed at the Duke Center for In Vivo Microscopy, an NIH/NIBIB national Biomedical Technology Resource Center (P41 EB015897), with additional support from NCI (U24 CA092656). Liposomal contrast agent was provided by Ketan Ghaghada and Ananth Annapragada (Texas Children's Hospital). We thank Sally Zimney for editorial assistance, Lucy Upchurch for assistance in preparing the supplemental material, and Brian Avants for helpful discussions regarding the use of ANTs.




home | about us | staff | contact | Duke Med Center