Optimizing photon counting detector CT reconstructions for automatic vessel tracing in prostate artery embolization

Prostate artery embolization (PAE) requires precise visualization of complex pelvic arterial anatomy and can be aided by advanced vessel navigation tools. Photon counting detector (PCD) CT provides simultaneous ultra-high resolution and material-specific imaging, potentially improving automatic vessel tracing workflow for procedural planning and guidance.

To identify, within a PCD-CT workflow, the reconstruction that provides the most robust and consistent automatic vessel tracing for pre-procedural PAE planning.

Ten patients underwent PCD-CT angiography of the prostate arteries in ultra-high-resolution mode. Four reconstruction approaches were evaluated: conventional reconstructions with (1) standard of care kernel and (2) sharper kernel (Bv64), along with (3) monoenergetic 50 keV, and (4) iodine map spectral results. Using commercial embolization guidance software, three vessels were manually traced for each patient to establish ground truth segmentation. Subsequently, automatic vessel traces were generated using identical start/end points as the ground truth. Root mean squared error (RMSE) between automatic traces and ground truth was used to quantify tracing accuracy.

Comparing individual traces, iodine maps demonstrated consistency with median RMSE of 0.7 mm (0.4 mm interquartile range (IQR)) versus 0.7 mm (12.5 mm IQR), 1.5 mm (14.7 mm IQR), and 0.8 mm (6.7 mm IQR) for Bv56, Bv64, and 50 keV reconstructions, respectively. Iodine maps produced fewer outliers, with 13.3% of traces and 40.0% of patients having one or more traces exceeding 5 mm RMSE, compared with 26.7%-36.7% of traces and 60.0%-80.0% of patients for the other reconstructions.

Material-specific iodine reconstructions from PCD-CT yield more robust automatic vessel tracing for PAE planning, offering a workflow-compatible solution requiring minimal protocol modifications.