以擴散磁振造影評估活體人腦之微觀結構

Abstract

隨著醫療技術的日新月異，諸多疾病例如缺血性心臟病與中風，其致病率與死亡率都獲得了極大的改善；然而，與精神疾病相關的失能與自殺率，和過去相比卻無明顯的下降；因此，我們對於人腦必須有更深入而徹底的瞭解。對活體人腦進行評估一直是一件非常困難而富有挑戰性的工作，為了解開人腦的謎團，核磁共振是最常使用的影像工具之一。 擴散頻譜磁振造影(Diffusion Spectrum Imaging, DSI)是一種評估人腦細微結構的先進技術，它能用來評估腦中的神經纖維走向與不等向性分率。除此之外，它在q空間中的信號可能還傳達了其他重要的生物參數，例如神經纖維的水分子分率(Axonal Water Fraction, AWF)。目前已經有很多的生物模型涵蓋了AWF的測量，例如雙指數模型(Bi-Exponential model)、擴散峰度磁振造影(Diffusional Kurtosis Imaging, DKI)、ActiveAx模型與NODDI模型。因為胼胝體的神經纖維走向相當一致，為了證實生物模型的可靠性，我們將針對人腦的胼胝體進行評估。 在我們的第一個實驗中，我們收集了一個相當特別的DSI數據，它在q空間與影像空間中都是二維的數據。將改良過後的ActiveAx模型應用到我們的二維DSI數據之後，我們可以估計人腦胼胝體不同部位的AWF。實驗結果與Aboitiz的組織學觀察相當一致，證實了實驗設計的可行性。在第二個實驗中，我們藉由事先建構好的NTU-DSI-122模板，另外建構了一個DKI模板。我們選擇了限制性線性最小平方法(Constrained Linear Least Square)來計算峰度張量(kurtosis tensor)的元素；另外，藉由DKI 的組織模型，我們能夠在較低的電腦計算負擔下，畫出全腦的AWF圖譜。總結來說，我們的研究能夠增進後續研究者對於人腦細微結構的探索，並在未來應用於精神或神經疾病的診斷，例如多發性硬化症(Multiple Sclerosis)的早期發現。

Although morbidity and mortality rate of some diseases, such as ischemic cardiac attack or stroke, decrease prominently with improvement in medical technology, disability and suicide rate related to mental diseases remain as before. Therefore, it is crucial for us to achieve a better comprehension of human brains. However, in vivo evaluation of human brains is always difficult and challenging. Magnetic Resonance Imaging (MRI) is the most commonly used method to tackle the mystery of human brains. Diffusion Spectrum Imaging (DSI), a kind of diffusion MRI, is an advanced technique for evaluating microstructure of the human brain, such as axonal direction and general fractional anisotropy (GFA). Its signal in the q space may convey other important biological parameters such as axonal water fraction (AWF). There are lots of researchers who integrate AWF measurement into their models, including Bi-Exponential model, Diffusional Kurtosis Imaging (DKI), CHARMED model, ActiveAx model, and NODDI model. To confirm the reliability of various models, we focus on AWF of the human corpus callosum (CC), in which axonal directions are very coherent. In our first experiment, we acquire a special DSI data, which is 2D in q space as well as in image space. Applying a modified ActiveAx model to our 2D DSI data, we can estimate AWF over the human CC. Our result is consistent to Aboitiz’s observation in histology, demonstrating the feasibility of our experimental setup. In the second experiment, we construct a DKI template from NTU-DSI-122 template, which is a DSI template. Constrained Linear Least Square (CLLS) is chosen for calculating kurtosis tensor elements. In addition, using a tissue model for DKI, we are able to obtain AWF map of whole brain with low computational loads. In conclusion, the proposed methods can facilitate further researches in probing microstructure of the human brain, and hence contribute to early diagnosis of mental or neurological diseases such as multiple sclerosis.