Program Information
A Computational Phantom of the Thorax Combining Anatomical and Respiratory Motion Models: Feasibility Study and Preliminary Developments
M Seregni1, M Riboldi1 , G Baroni1,2 , G Sharp3* , (1) Politecnico di Milano, Milano, Italy, (2) Bioengineering Unit, Centro Nazionale di Adroterapia Oncologica, Pavia, Italy, (3) Massachusetts General Hospital, Boston, MA
Presentations
SU-K-201-15 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 201
Purpose: To develop an open-source computational phantom of the thorax supporting lung cancer imaging and treatment. Specifically, we report the preliminary development of anatomical and respiratory motion models embedded in the phantom.
Methods: We considered six publically available 4DCTs of lung cancer patients acquired at CLB (Lyon, France). First, the end-exhale (EE) volumes of each patient (Ip) were rigidly aligned to the EE volume of an arbitrarily selected patient, then the aligned volumes were averaged creating an initial reference image (I0). Second, we warped each Ip to I0 through a B-spline deformation vector field (DVF). The average of the warped volumes became the updated reference image for the next iteration. By applying such a procedure iteratively, we improved progressively the quality of the average EE volume, which represents the final anatomy of the phantom.For each patient, we registered the end-inhale, mid-inhale and mid-exhale volumes to EE using B-splines. Then, each patient-specific motion was warped to the phantom anatomy. The average motion and its inter-patient variability were calculated using PCA. Finally, a model representing the average motion (AV) and the average plus or minus one standard deviation (AV+SD, AV-SD) were applied to the phantom EE anatomy.
Results: Visual inspection of the phantom anatomy shows that high-contrast structures and interfaces (spine, lungs, diaphragm, chest wall) can be clearly identified. However, blurring affects lower-contrast structures. The registration accuracy was within 1.4±1.5mm (median±IQR, worst case). The phantom EE lung volume was 3400cc. End-inhale volumes were 3711, 3748, 3652cc considering the AV, AV+SD and AV-SD motion, respectively.
Conclusion: We created an anatomic phantom prototype consisting of a population-based model of the thoracic anatomy and respiratory motion. Future work will be focused on including a statistical model of lung tumor volumes and on improving the quality of the anatomy by using a larger patient population.
Contact Email: