Development of a radiology-based computational analysis to evaluate osteointegration of a xenohybrid bone substitute

The main goal of this master’s degree thesis is to develop a reproducible analysis method which, starting from a set of CT images of patients who underwent surgical procedures involving SmartBone® grafts, could provide a time-related performance index of the material. The presented work needs to be conceived as a progression of a former master’s degree thesis in which a radiological evaluation method for the osteointegration of SmartBone® has been developed. To try to overcome the difficulties encountered with the former method, another three-dimensional approach has been proposed. This analysis aims to obtain an accurate 3D model of SmartBone® from the patient’s CT image stack, that allows to extract two crucial parameters: graft volume and its average HU value, which could represent two useful indicators of how SmartBone® interact with natural bone tissue. Two software have been used: ITK-SNAP and Inobitec Dicom Viewer that allow to obtain three-dimensional volume segmentations applying region-growing algorithms. This approach offers the possibility for the operator to dynamically modify the segmented structure to correct any algorithm inaccuracy and to real-time visualize the 3D rendered volume. Being able to generate a valid 3D model of the implanted material for every radiological follow-up means monitoring the structural changes covered by the graft during time. The average HU value offers an index of osteointegration of the implant, describing the host’s response to the external material. The objective of the procedure is to identify SmartBone® as precisely as possible and to do this, the implant must be easily distinguishable from the surrounding tissue. However, this is not always possible due to scanning parameters and the advanced degree of reabsorption of the bone graft. Whenever a precise 3D segmentation of SmartBone® in not possible, a 2D analysis represents an alternative procedure. In this case, the average HU value needs to be considered as the main performance index. In this respect, the aim of this 2D analysis in to evaluate how this parameter changes over time in the same reference area. Being able to precisely identify the same anatomical area for both follow-up exams could be a challenging task, therefore the registration tool implemented in ITK-SNAP has been used, which represents a solid support to easily align different data taken at different time steps by simply rotating and translating the image stacks. The analysis procedure proposes to segment SmartBone® in the first dataset, in which it is clearly visible, and to import the obtained voxel map in the second stack. Thus, the implant interaction with surrounding tissue can be measured observing changes in the average HU value of the two selected ROIs in both follow-up exams. On the other hand, this 2D analysis does not offer information about structural modifications of the implant, therefore the 3D analysis, whenever feasible, is preferable. To validate the here proposed evaluation method, four clinical cases have been considered. A complete analysis has been done for two orthopedic cases, in which SmartBone® is often used in blocks and for two cranio-maxillo-facial cases, in which the implant is frequently present in form of microchips or plates. In this way, it was possible to evaluate the efficiency of the analysis method apart from the implant’s geometry and the clinical application.