| 摘要: | 核磁共振技術(Magnetic Resonance Imaging, MRI)為現今臨床上重要的檢測技術，其核磁共振技術最大優點是對人體不具侵襲性，且可以多方向掃描，並提供三度空間、高對比度的影像。可有利於醫師對疾病的診斷更加準確，以提高治療的正面效果。當一病人被安排做MRI造影後，對人體某一器官切面部位，會產生一系列的多頻譜影像(multispectral image)。如果把一系列的切面影像疊在一起組合而成後，便形成人體一個立體的三維結構。醫師就藉由這個立體結構得到一些醫學診斷的資訊，如器官的形狀、位置與體積大小。雖然經由多頻譜影像可獲得更多的資訊，但也造成病理判讀上的困擾。因此，我們將這些多頻譜影像經過精準的轉換法處理後，形成單一強化組織影像讓醫生更容易的對病理做診斷。此篇論文提出了一個新特徵自我選取的方法，Target Generation Process(TGP)。並將TGP合併於Linear Discriminant Analysis (LDA)、Constrained Energy Minimization (CEM) filter 與 Fuzzy Knowledge Based Seeded Region Growing (FKSRG)三個方法中。其中我們稱此前兩方法為TGP Linear Discriminant Analysis (TGP+LDA) 與 TGP Constrained Energy Minimization (TGP+CEM)。而TGP可幫助FKSRG改善區域合併(Regions Merging)時的不確定性，利用這些方法來強化出腦中的CSF(Cerebrospinal Fluid)、白質(White Matter)以及灰質(Gray Matter)三大組織，使醫生做診斷時更加有效率。因此我們的工作即在研究如何從多頻譜MRI影像中，將腦部的主要組織(如CSF、WM、GM)給強化出來，更進一步的發展是可將腦腫瘤明確的強化與切割出來。最後我們將加入常見的FMRIB’s Automated Segmentation Tool (FAST) 、 Fuzzy C-means (FCM) 與 C-means (CM)等方法來進行比較，並且提出一套方法來評比這些方法的可行性與強健性
Magnetic Resonance Imaging (MRI) is a useful medical instrument in medical science. It provides unparallel capability of revealing soft tissue characterization as well as 3-D visualization and proposes the diagnosis without needing to intrude into the human body. MRI produces a sequence of multiple spectral images of tissues with a variety of contrasts, but the multispectral images cannot be conveniently used to be a pathology diagnosis correctly. In general, we need to transform the multispectral images to an enhanced image, which is easier to be used for doctor’s clinical diagnosis. One of the potential applications of MRI in clinical practice is the brain parenchyma classification. In this dissertation, we present an automatic feature selection method called Target Generation Process (TGP) and combine it with Linear Discriminant Analysis (LDA), Constrained Energy Minimization (CEM) filter and Fuzzy Knowledge Based Seeded Region Growing (FKSRG) to automatically enhance, classify and segment the three major tissues, i.e. gray matter (GM), white matter (WM) and cerebral spinal fluid (CSF), from a multispectral MR image of the human brain. The TGP is a fuzzy-set process that generates a set of potential targets from unknown information and then the targets are applied into LDA, CEM and FKSRG methods. It can support LDA and CEM to become the TGP+LDA, TGP+CEM filter. In conventional regions merging of Seeded Region Growing (SRG) algorithm, the final number of regions is unknown. Therefore, TGP is proposed and applied to support conventional regions merging, such that the FKSRG method does not produce over or under segment images. Finally, two images sets, namely, computer-generated phantom images and real MR images, are used in experiments to assess the effectiveness of the proposed methods. Experimental results demonstrate that proposed methods segment multispectral MR images much more effectively than the FMRIB’s Automated Segmentation Tool (FAST), Fuzzy C-means (FCM) and C-means (CM) methods. |
