Quantification of metal artifacts in metal artifact avoidance

The use cone beam computed tomography (CBCT) imaging allows for a more precise placement of transpedicular screws during spine surgery, but can suffer from the emergence of metal artifacts. While approaches exist, that successfully mitigate those artifacts and improve the image quality by optimizing the trajectory of the C-arm, they do not give insight in how much the image quality changes between different candidate trajectories and whether the image quality will be sufficient. To establish this connection, we introduce a pipeline to directly predict the quality of an image given only a set of scout views and candidate trajectory. A set of metrics was created to quantify the loss in image quality due the metal artifacts. Additionally, a set of predictors was investigated concerning their capability to predict the results of the metrics for a scan based on metal path lengths within the candidate trajectory. A phantom was created to evaluate both the metrics and the predictors in a controlled setting. The results of the introduced metrics successfully mirror the underlying metal artifacts. Both the contrast-to-noise ratio and inner contrast metric show the expected behavior of the image quality increasing at higher angles between metal and detector, while the outlier metric shows an inverse behavior. The different predictor versions show good results when tested for linear correlation to the introduced metrics. The results show that it is generally possible to predict the quality of and image based on the utilized C-arm trajectory. By extending the model with additional external factors and validating it with expert knowledge, it has the potential to significantly improve the clinical the workflow of if image-guided spine surgery.