Hybrid Deformable Image Registration for Advanced Applications in Image-guided Interventions
Navid Samavati · TSpace (University of Toronto) · 2015
Accounting for tissue deformation in medical imaging through Deformable Image Registration (DIR) is of significant clinical importance in diagnostic and therapeutic applications. Recent advancements in Magnetic Resonance Imaging (MRI) have enabled better diagnosis of cancerous lesions in soft tissue organs such the prostate. However, correlative pathology is needed to validate the precise boundaries of MRI diagnostic findings. Correlation of histopathology with in vivo imaging involves several mechanical deformations to the prostate tissue, demanding the use of a DIR. At the therapy level, tissue deformations and physiological motion can lead to significant differences between the planned and the delivered radiation dose in radiotherapy treatments such as the lung. Deformable dose reconstruction using an accurate DIR is essential in better estimation of the delivered dose at the time of planning. A hybrid biomechanical intensity-based technique was developed and validated using 4-dimensional Computed Tomography (4DCT) of lung cancer patients. The hybrid method was designed to take advantage of the benefits of the intensity-based algorithms (i.e. use of high contrast structures for accurate alignment) without losing the physical plausibility of complex physiological deformations obtained through the biomechanical modeling process. The combination of biomechanical and intensity-based registrations enabled a significantly more accurate registration with an average of 1.5 mm compared to 3.1 mm and 2.6 mm using each of the methods alone, respectively. The method was further tested using prostate MRI with average improvements of 0.7 mm compared to the biomechanical DIR. The biomechanical model based DIR was adapted for the correlation of histopathology with MRI showing the limited (53%) sensitivity of MR imaging in detecting the true cancerous regions in the prostate. This result is fundamental to the delivery of various treatments for locally recurrent prostate cancer where accurate definition of tumor burden is required. The impact of the hybrid method in radiation therapy was assessed in a retrospective lung dose calculation study. The results revealed that 1.6 mm reductions of average registration error using the hybrid method translates to over 1 Gy dose difference in up to 50% of a lung radiation therapy patient population, which may be clinically significant. In summary, the work in this thesis resulted in the development of a highly accurate hybrid DIR technique with substantial effects on understanding the prostate cancer diagnosis sensitivity, and the dosimetric calculations for lung radiotherapy.