MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Region-Of-Interest Based Local Principal Component Analysis Algorithm to Extract Independent Motion Pattern From Cone-Beam Tomography Projections

P Tsai*, B Lu , G Yan , J Park , J Wu , C Liu , university Florida, Gainesville, FL

Presentations

(Sunday, 7/29/2018) 4:00 PM - 4:30 PM

Room: Exhibit Hall | Forum 6

Purpose: The local principal component analysis (LPCA) method, is able to extract patient’s respiratory motion pattern using cone-beam tomography (CBCT) projections. However, it lacks the capability to separate useful motion signals between tumors and the diaphragm. In addition, the detection accuracy of LPCA is largely affected by the perturbation of outliers. We developed a region-of-interest based LPCA (ROI-LPCA) algorithm, which can help on both drawbacks of LPCA method.

Methods: Unlike the original LPCA method considering all images information as a whole, we separate ROIs enclosing only the tumor or the diaphragm. The trajectory of the tracking ROI on CBCT projection was then automatically calculated based on its geometrical information. Later, pixel values from each projection were summed to form ROI based Amsterdam Shroud (AS) images. Adaptive z-normalization was applied to the AS images to enhance contrast prior to ROI-LPCA. A CIRS dynamic phantom with 2.5cm diameter sphere and a surrogate platform was used to simulate independent motion patterns for tumor and diaphragm. An optical tracking system (OTS) was employed to obtain the ground truth as the reference. The performance of the algorithm was evaluated by comparing the extracted peak locations and the average cycle with the reference.

Results: For tumor motion, the peak locations extracted with ROI-LPCA and LPCA compared with the reference had 100% and 41.46% match within one projection, respectively; for the surrogate motion, 35 out of 41 peak locations matched with the reference within one projection. The average cycle of extracted motions deviated from the reference was within 0.018 seconds.

Conclusion: A ROI-LPCA algorithm capable of tracking individual ROIs was developed. It eliminated the impact of image outliers. The extracted motion patterns had sufficient accuracy. It enables us to track multiple independent ROIs for statistical dependence analysis prior to treatment as a pattern detection and verification tool.

Keywords

Cone-beam CT, Statistical Analysis, Treatment Verification

Taxonomy

IM/TH- RT X-ray Imaging: CBCT imaging/therapy implementation

Contact Email