M. Bashoor Zadeh, M. Bohner, G. Baroud (Sherbrooke, Canada-Bern, Switzerland)
INTRODUCTION:
Micro computed tomography (μCT) is a powerful tool for non invasive quantification of bone substitute structure. However, image resolution as an acquisition parameter has a significant impact on accuracy of structural analysis. Particularly when the voxel size is larger than the structure of interest, it is difficult to obtain accurate morphological information. This study applies an artificial subvoxel processing method presented in [1] to increase the image resolution, leading to high precision of morphological measurement. A good consistency was found between the results obtained from artificial method and actual high resolution scanning.
METHODS:
Two cylindrical samples of β-TCP bone substitute with similar geometric features were scanned by μCT system with 15 and 30 μm resolutions. The low resolution scans (30 μm) were then subdivided into small voxels by using an artificial subvoxel process algorithm called “single-pass subvoxel processing” [1]. According to “single-pass subvoxel processing” algorithm the allotted intensity of each subvoxel is determined by local neighbouring criteria and strict conservation principles. Due to similarity of grey level histogram before and after subvoxel process, the same threshold values were used in both cases. Geometrical parameters were defined based on 3D fuzzy distance transform (FDT) map [2]. Where, average pore size and interconnection size are equal to the average FDT values of local maxima and saddle points in FDT map, respectively.
RESULTS:
Fig. 1 shows the pore size and interconnection size distributions of one sample based on different resolutions and methods. As illustrated in Fig.1-a, after decreasing the voxel size, the distributions of pore didn’t change significantly. Conversely, interconnection size distributions show significant difference in the number of interconnections with larger size (Fig. 1-b).
The mean and standard deviation for each parameter as a function of voxel size and method are summarized in table 1. All parameters demonstrate the similar pattern of change for both methods (actual scanning and artificial subvoxel process).
Detecting small struts in high resolution images results in smaller pores among the sample and so decreasing the average pore size. Also modification of interface surface of pores in images provided by small voxel size leads to finding larger interconnections. This can be a reason for increasing the average interconnection size.
Fig.1- (a) Pore size distribution, (b) Interconnection size distribution.
Voxel size
(µm) Method Threshold values Porosity (%) Pore size (µm) Inter. size (µm)
30 Scan 50-110 52 ± 1 886 ± 32 109 ± 13
15 Scan 30-80 53 ± 2 879 ± 27 151 ± 4
15 Artificial 50-110 53 ± 1 857 ± 47 217 ± 25
Table1- Average and standard deviation of geometrical parameters for each voxel size and method, (Inter. = interconnection).
CONCLUSION:
the artificial subvoxel processing method demonstrates to be a useful image processing tool for more accurate quantification of bone substitute structure.
ACKNOWLEDGMENT:
The authors thank T. Vincent for support provided for the subvoxel processing code.
REFERENCES: 1. Hwang SN, Wehrli FW. Magm Reson Med, 47(2002)948-957.
2. Bashoor Zadeh M, et al. Tissue Eng, 13(2007)1377.