頭頸部磁振血管造影影像校準及融合

Image Registration and Fusion in 3D-TOF and Contrast Enhanced Magnetic Resonance Angiographies of Head and Neck

磁振造影血管影像(Magnetic Resonance Angiography, MRA)為現階段診斷腦血管疾病的主要工具之一，目前較常使用之技術包括為3D-TOF (Time-of-flight)磁振血管造影與注射顯影劑之磁振造影血管造影。3D TOF之磁振血管影像解析度較高，主要以呈現顱內血管遠端頸動脈和脊椎動脈影像為主，診斷範圍較小無法涵蓋全部頸動脈。注射顯影劑之磁振造影血管影像，可完整呈現兩側頸脊椎動脈與顱內血管；但其解析度較低。本研究針對注射顯影劑磁振造影血管影像和3D TOF磁振血管影像進行三維校準融合。並利用粒子群最佳化演算法(Particle Swarm Optimization, PSO)與基因演算法(Genetic Algorithm, GA)搜尋校準之最佳幾何轉換參數組合。實驗結果顯示，粒子群最佳化演算法無論在校準最小誤差及執行時間上皆優於基因演算法。本研究藉由影像處理和影像三維自動校準技術開發電腦輔助診斷系統，可呈現較單一來源更為完整之頭頸部血管三維資訊，以提昇臨床診斷之醫療品質。

3D-T0F and contrast enhanced Magnetic Resonance Angiographies (MRA's) are the most commonly used tools for diagnosis of cerebrovascular diseases. The 3D-TOF image acquisition techniques are applied to acquire intracranial vascular; meanwhile, the contrast enhancement image acquisition techniques are used to acquire bilateral internal carotid artery. Due to the sampling constraints, the contrast enhanced MRA images results in the coarse resolution. On the contrary, the 3D-TOF images present a small range of intracranial vascular imaging. In this paper, the axial plane MRA and the coronal plane MRA from various sources are fused to provide more comprehensive images than each one of them. Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) are applied to search for the optimum of geometric parameters. The results of performance comparison showed that PSO outperformed GA both in alignment quality and computational time. In this paper, image registration tools for fusing 3D information of cerebrovascular and carotid from various sources of MRA images are developed to enhance the quality of medical research and clinical diagnosis.