Loading...
Optical Flow Based Extraction of Breathing Signal from Cone-Beam CT Projections
Sabah, Shafiya
Sabah, Shafiya
Description
A Master of Science thesis in Biomedical Engineering by Shafiya Sabah entitled, “Optical Flow Based Extraction of Breathing Signal from Cone-Beam CT Projections”, submitted in May 2020. Thesis advisor is Dr. Salam Dhou. Soft copy is available (Thesis, Approval Signatures, Completion Certificate, and AUS Archives Consent Form).
Abstract
Lung cancer continues to be the most common type of cancer worldwide. Radiotherapy is used to break tumor cells by application of radiation beams during cancer treatment. Adapting radiotherapy to respiratory movements has always been a major concern in thoracic and upper-abdomen radiotherapy. Thus, estimating the respiration induced organ motion has been widely studied. Respiratory motion can be estimated using external equipment that trace the chest motion during breathing or tracking of radio-opaque markers implanted surgically in the lungs. However, these techniques are either invasive or require external equipment. The objective of this thesis is to propose an image-based method to estimate the respiratory motion without the involvement of any external equipment or implanted markers. To achieve this objective optical flow was implemented to acquire dense motion vectors from sequence of cone beam CT projection images. Principal component analysis was then performed on the motion vectors to project data to a lower dimension while preserving the dataset motion trends. Extracted signal was sorted into phase bins followed by 4D-reconstruction using a single phase to eliminate the effect of breathing motion. Several experiments were conducted to gauge the feasibility of this study. The dataset comprised of three computer simulations of cone beam CT projections as well as three real datasets acquired under clinical settings. The computer simulated phantom datasets were generated to include different scenarios such as fast breathing, slow breathing and irregular breathing patterns. The average phase shift error for phantom dataset under fast breathing motion was 0.6 ± 0.66 projections, 0.4 ± 0.5 under slow breathing and 0.53 ± 0.51 under irregular breathing. For clinical datasets the average phase shift for patient 1 was observed to be 1.936 ± 0.734, 1.185 ± 0.781 for patient 2 and 1.537 ± 0.93 for patient 3. Four-dimensional CBCT reconstruction was performed for one phantom dataset. Reconstructed image had a peak signal to noise ratio of 45.75.