A new cerebral vessel benchmark dataset (CAPUT) for validation of image-based aneurysm deformation estimation algorithms.

Hemodynamic properties and deformation of vessel structures are assumed to be correlated to the initiation, development, and rupture of cerebral aneurysms. Therefore, precise quantification of wall motion is essential. However, using standard-of-care imaging data, approaches for patient-specific estimation of pulsatile deformation are prone to uncertainties due to, e.g., contrast agent inflow-related intensity changes and small deformation compared to the image resolution. A ground truth dataset that allows evaluating and finetuning algorithms for deformation estimation is lacking. We designed a flow phantom with deformable structures that resemble cerebral vessels and exhibit physiologically plausible deformation. The deformation was simultaneously recorded using a flat panel CT and a video camera, yielding video data with higher resolution and SNR, which was used to compute 'ground truth' structure deformation measures. The dataset was further applied to evaluate registration-based deformation estimation. The results illustrate that registration approaches can be used to estimate deformation with adequate precision. Yet, the accuracy depended on the registration parameters, illustrating the need to evaluate and finetune deformation estimation approaches by ground truth data. To fill the existing gap, the acquired benchmark dataset is provided freely available as the CAPUT (Cerebral Aneurysm PUlsation Testing) dataset, accessible at https://www.github.com/IPMI-ICNS-UKE/CAPUT .