7.1
Conclusions
In this thesis, X0ray µCT was used as the primary tool for analysis of porous SLM Ti structures and the bone ingrowth into porous Ti and bioactive glass implants in a rat tibia model. A novel quantification approach was developed, involving the creation of an average model from within each
in vivo experiment that aids objective VOI location and orientation in different animals. The average
model was applied to two separate case studies, implantation of a porous 70S30C bioactive glass that dissolves over time inside the body and the bone ingrowth into porous Ti implants over time. Novel segmentation algorithms were also developed during the course of the thesis to improve the accuracy of image analysis.
7.1.1
X"ray µCT as a quality control tool
The direct comparison and tracking of randomised Ti porous implants over a series of cleaning procedures was demonstrated for the first time, highlighting the evolution of the complex Ti structure. The combined use of full and local tomography was also demonstrated, to provide more detailed
information about the surface whilst not sacrificing information of the bulk structure. A number of key observations were made to further improve the design of SLM Ti implants:
1) Low angle struts were poorly connected and not as strong. These can be avoided during the randomisation process and removed in future designs.
2) Bending and failure of un0terminated struts at the surface of Ti structures as observed during cleaning. A simple solution was suggested: designing all the struts at the surface to be node0 terminating, providing greater stability.
3) The pore and strut sizes were measured using an accessible volume code. Built implants can be verified to initial CAD designs quickly by µCT.
Using local tomography techniques limits the effective field of view during reconstruction. This is especially limiting when the detector is small (512 pixels) as it was in my case leading to only a very small area being analysed at high resolution. The technique has limiting factors such as increased noise and potentially fluctuating X0ray attenuations, but it was necessary in order to obtain high resolution scans.
The registration techniques demonstrated in Chapter 4 were extremely useful for any future work involving the registration of animal limbs. This chapter showed that µCT was capable of producing datasets with enough information and with suitable resolution for registration. This is a huge advantage for µCT, as this is not possible using other techniques. Although it was not implemented in these studies, it would be possible to pre0scan a porous implant then compare it periodically whilst the implant was still inside an animal (making sure the exposure of X0rays to the animal was not influencing the animal’s normal growth or development).
7.1.2
Average model as an essential analysis procedure
The average tibial model was developed as a response to the arbitrary and incompatible methods currently employed to determine a VOI for quantification. The impact of the approach is that it is applicable to different species, different animals as the average model is derived directly from the animal model being used in the experiments. It provides an automatic and objective way of determining the VOI minimising the element of human error and bias. This leads to quantification results that are normalised by each unique VOI, allowing the direct comparison of results between groups. The use of the contralateral limb to create the average model also reduces the number of animals that need to be sacrificed. The quantification results from using this approach was also verified by histological quantification in (Midha et al., 2013a).
In2vivo and ex2vivo implant studies in animal models are still highly dependent on the skill of the
surgeon. In the early surgeries carried out for use in this thesis, there were several issues that could have affected the quantification results. First of all, implant orientation is extremely difficult to standardise especially when handling such small implants (3 mm diameter) and being press0fit. There is also a tendency for implants to be pushed out (by the blood pressure) or be dislodged early on as the animals move about freely. This not only leads to the implant not spanning the entire defect area but also vastly differing VOI sizes, that may no longer be studying the same effect. In follow0on studies, this problem was avoided by having a thin strand of Ti that allowed greater control on placement of the implant in the defect space and which could then be broken off after implantation. Secondly, there were a number of fractures (on average 1 in each group) in the tibia that led to a greater callus response which covered the entire leg rather than the defect area and bone forming across the fracture to join the two parts of the limb together. This limited the number of animals that could be successfully analysed. Another factor in the quantification is the number of contralateral limbs used to create the average model. Ideally the model should be representative of the animal sample set being used. Currently there is further work going on to finding what number of limbs, at
what resolution is acceptable to be representative. This work will hopefully further refine the method to produce better models.
The average model was applied to bone ingrowth into 70S30C and Ti implant materials, from which it was concluded:
1) Preconditioning led to the increased dissolution of 70S30C scaffolds. This significantly improved the bone growth into the scaffold over time in terms of bone volume ingrowth. 2) Not preconditioning 70S30C scaffolds inhibited the growth of bone into the scaffold pores
and led to a curvature in the final healed bone. This was reflected in a new parameter, ⁄
3) Actifuse and NovaBone were successful osteoconductors, but caused large bone growth inside the marrow, filling up to 50% of the marrow volume.
4) Preconditioned 70S30C exhibited bone ingrowth values similar to commercially available products, making preconditioning of sol0gel scaffolds a pre0requisite before implantation. This also led to developments of new compositions that reduced the Ca content in the glasses or compositions that balanced local pH by additional dissolution components.
5) µCT analysis of Ti implants over time showed significant changes in BIC, , SSA and bone thickness, but no differences were found between dry Ti, Ti+PRP and Ti+blood cases. This suggested that Ti does not hinder bone ingrowth.
Whilst X0ray µCT is clearly a powerful imaging tool, it has yet to fully be utilised as a quantification tool. There are still some limitations to µCT that need to be overcome, namely resolution and bias. With µCT machines becoming more and more powerful, improved resolution and more sensitive detectors are allowing sub0micron imaging. On improving the subjectivity of quantification values, there are numerous attempts to automate segmentation steps, to eliminate user interaction. This thesis has presented an objective method of finding the VOI, a critical step in quantification of 3D volumes, which is not only applicable in long bone defects but other bony sites.