Demonstrate consistent measured image-quality gains

Demonstrate consistent image-quality gains from continuous sparse-view cone-beam CT trajectory optimisation in physical robot-based acquisitions relative to uniform sampling on the same reachable acquisition band.

Background

The paper develops a gradient-based method that continuously optimises individual X-ray source poses on several scanner kinematic manifolds, combining soft-Tuy coverage, an attenuation-aware bundle penalty, and a continuous View Covariance Loss criterion. Virtual experiments show substantial gains on the asymmetric Defrise flange and competitive performance relative to discrete view-selection methods.

A robot-based camera experiment establishes that the optimised pose sets can be physically acquired and reconstructed in a common reference frame. However, the measured results do not show a consistent metric improvement over uniform sampling within the same elevation band: uniform sampling outperforms the planned arms for several budgets and metrics, while reference-acquisition artefacts and independently acquired sessions complicate evaluation. The authors therefore leave consistent measured image-quality improvement unresolved.