|Title:||Investigation of the XCAT phantom as a validation tool in cardiac MRI tracking algorithms.|
|Citation:||Phys Med. 2018 Jan;45:44-51.|
|Abstract:||PURPOSE: To describe our magnetic resonance imaging (MRI) simulated implementation of the 4D digital extended cardio torso (XCAT) phantom to validate our previously developed cardiac tracking techniques. Real-time tracking will play an important role in the non-invasive treatment of atrial fibrillation with MRI-guided radiosurgery. In addition, to show how quantifiable measures of tracking accuracy and patient-specific physiology could influence MRI tracking algorithm design. METHODS: Twenty virtual patients were subjected to simulated MRI scans that closely model the proposed real-world scenario to allow verification of the tracking technique's algorithm. The generated phantoms provide ground-truth motions which were compared to the target motions output from our tracking algorithm. The patient-specific tracking error, ep, was the 3D difference (vector length) between the ground-truth and algorithm trajectories. The tracking errors of two combinations of new tracking algorithm functions that were anticipated to improve tracking accuracy were studied. Additionally, the correlation of key physiological parameters with tracking accuracy was investigated. RESULTS: Our original cardiac tracking algorithm resulted in a mean tracking error of 3.7 ± 0.6 mm over all virtual patients. The two combinations of tracking functions demonstrated comparable mean tracking errors however indicating that the optimal tracking algorithm may be patient-specific. CONCLUSIONS: Current and future MRI tracking strategies are likely to benefit from this virtual validation method since no time-resolved 4D ground-truth signal can currently be derived from purely image-based studies.|
|Type of Work:||Article|
|Type of Publication:||Post-print|
|Appears in Collections:||Research Papers and Publications. Sydney Medical School|
|Lowther_PM_2018.pdf||675.14 kB||Adobe PDF|
Items in Sydney eScholarship Repository are protected by copyright, with all rights reserved, unless otherwise indicated.