The objectives of the project were fully satisfied. The crux of the problem was to find a relationship between the trackers and the bones of the spine. To find this mapping our group requires only one CT scan to be taken prior to surgery. From there, a single program is run, which is a combination of the TCA, the Functional BCA, and the Transformation Calculator. This single program needs the CT scan, and a few simple user inputs, and the mapping between bones and trackers will be returned (as described in the ‘Methods’ section).
Going forward, the next natural step would be to integrate this transformation matrix with the NDI Camera software. Since the program produces a 1-1 linear relationship between the trackers and the bones, to integrate our program with the camera’s software, the only required step is to apply the linear transformation to the real-time values the camera reads. Although a mathematically simple task, this step requires familiarity with the technical specifications and backend of the NDI software. To achieve this, the real-time values must be extracted from the software, such that the transformation could be applied. This finalizing step allows the user to determine all critical measurements within the NDI software.
Our team is very confident in the robustness of the current TCA. This portion of the program runs without error, and without outliers. However, as described in the section for the BCA, we see potential for improvement. Although the functionality of our current BCA is not compromised, some error is introduced when the user is prompted to select points to parameterize the bones. Since all doctors use a similar type of manual plotting to validate the spine procedure’s successfulness, our team finds no issue using this feature to parameterize the bones. However, if the 3D V-Net Machine Learning Algorithm is adopted correctly, this could allow the user to simply select the name of the bone they wish to measure. This feature would immediately reduce the time it takes to run the program. With that, it would eliminate the human error introduced by plotting the points manually. For these reasons, our team agrees that the next iteration of this project should address this issue.