Feature-Oriented CBCT Self-Calibration Parameter Estimator for Arbitrary Trajectories: FORCAST-EST

Background: For the reconstruction of Cone-Beam CT volumes, exact position each projection is needed; however, in some situations, this information missing. Purpose: The development a self-calibration algorithm for arbitrary CBCT trajectories that does not need initial positions. Methods: Projections are simulated spherical grid around center rotation. Through using feature detection and matching, an acquired compared to image grid. with most matched features was used as starting point fine calibration state-of-the-art algorithm. Evaluation: This approach nearly correct positions when FORCASTER CMA-ES minimization normalized gradient (NGI) objective function. comparison metrics were root mean squared error, structural similarity index, dice coefficient, which evaluated on segmentation metal object. Results: parameter estimation regular 496 took 1:26 h following metric values: NRMSE = 0.0669; SSIM 0.992; NGI 0.75; Dice 0.96. 3:28 metrics: 0.0190; 0.999; 0.92; 0.99. 5:39 0.0037; 1.0; 0.98; 1.0. Conclusions: proposed can determine parameters orientations enough accuracy reconstruct 3D volume low errors.