MRI Measures of Aging
Methodological Issues
MRI 測量老化的方法論問題

Magnetic resonance imaging (MRI) is one of the most widely used imaging techniques in studies of cognitive neuroscience, including in brain aging. Compared to other medical imaging modalities, MRI has three important features that have contributed to this broad acceptance. First, MRI does not involve ionizing radiation and, as a matter of fact, most MRI scans do not require any exogenous contrast agents. Therefore, it is particularly suitable for studies on healthy participants in whom injection of exogenous tracer or agent is not desirable. For similar reasons, one can repeat the MRI scan as frequently as needed in a follow-up or longitudinal setting. Second, MRI provides an excellent spatial resolution, allowing the visualization of the brain with exceptional details. It also shows clear image contrast between soft tissues, unlike several other imaging technologies such as X-ray and computertomography (CT), thus can easily distinguish the gray and white matter based on their characteristic MR properties. Finally, MRI is also a versatile imaging modality. Within the same imaging session, one can perform several MRI pulse sequences and therefore obtain multiple domains of structural, functional, and physiological information from the participant. This advantage reduces subject burden and allows the integration of multi-parametric datasets in understanding cerebral aging. This chapter will provide a review of the basic principles of MRI, describe several major MRI techniques that are commonly used in cerebral aging, and introduce new, emerging techniques that are on the horizon.
磁共振成像（MRI）是認知神經科學研究中最廣泛使用的成像技術之一，包括大腦老化的研究。與其他醫學成像方式相比，MRI 有三個重要特徵促成了其廣泛的接受度。首先，MRI 不涉及電離輻射，事實上，大多數 MRI 掃描不需要任何外源性對比劑。因此，它特別適合於對健康參與者的研究，因為不希望注射外源性示蹤劑或藥劑。出於類似的原因，在後續或縱向研究中，可以根據需要頻繁重複 MRI 掃描。第二，MRI 提供了優秀的空間解析度，能夠以卓越的細節可視化大腦。它還顯示出軟組織之間的清晰影像對比，與其他幾種成像技術（如 X 光和電腦斷層掃描（CT））不同，因此可以輕鬆區分灰質和白質，基於其特徵性的 MRI 屬性。最後，MRI 也是一種多功能的成像方式。 在同一次影像檢查中，可以執行多個 MRI 脈衝序列，因此可以從參與者那裡獲得多個結構、功能和生理信息的領域。這一優勢減少了受試者的負擔，並允許在理解大腦老化時整合多參數數據集。本章將回顧 MRI 的基本原理，描述幾種在大腦老化中常用的主要 MRI 技術，並介紹一些即將出現的新技術。

The source of MRI signal is from the endogenous nuclei (e.g., ${}^1\mathrm{H},{}^{23}\mathrm{Na},{}^{31}\mathrm{P},{}^{17}\mathrm{O}$${}^1\mathrm{H},{}^{23}\mathrm{Na},{}^{31}\mathrm{P},{}^{17}\mathrm{O}$^(1)H,^(23)Na,^(31)P,^(17)O{ }^{1} \mathrm{H},{ }^{23} \mathrm{Na},{ }^{31} \mathrm{P},{ }^{17} \mathrm{O}, and ${}^{19}\mathrm{F}$${}^{19}\mathrm{F}$^(19)F{ }^{19} \mathrm{~F} ) that form the human body. The vast majority of the MRI studies used in cerebral aging are based on the detection of the protons in ${}^1\mathrm{H}$${}^1\mathrm{H}$^(1)H{ }^{1} \mathrm{H} nucleus (more specifically water protons); thus we will focus our discussion on this nuclear species.
MRI 信號的來源是來自於形成人體的內源性核（例如， ${}^1\mathrm{H},{}^{23}\mathrm{Na},{}^{31}\mathrm{P},{}^{17}\mathrm{O}$${}^1\mathrm{H},{}^{23}\mathrm{Na},{}^{31}\mathrm{P},{}^{17}\mathrm{O}$^(1)H,^(23)Na,^(31)P,^(17)O{ }^{1} \mathrm{H},{ }^{23} \mathrm{Na},{ }^{31} \mathrm{P},{ }^{17} \mathrm{O} 和 ${}^{19}\mathrm{F}$${}^{19}\mathrm{F}$^(19)F{ }^{19} \mathrm{~F} ）。大多數用於腦部老化的 MRI 研究都是基於對 ${}^1\mathrm{H}$${}^1\mathrm{H}$^(1)H{ }^{1} \mathrm{H} 核中的質子的檢測（更具體地說是水質子）；因此我們將重點討論這種核物種。

Each proton can be viewed as a small magnet. When these small magnets are placed in an environment without any significant external magnetic field (e.g., outside the MRI scanner), they will be oriented randomly (Figure 1.1A) and, as a result, their net strength, referred to as magnetization, is 0 . On the other hand, when these magnets are placed inside a strong external magnetic field (e.g., inside the MRI scanner), their orientation is no longer random. Slightly more than half of them will orient themselves to be parallel to the external field, while slightly less than half will be anti-parallel to the external field (Figure 1.1B). Thus, the net strength of the proton magnets is no longer 0 but is parallel to the external field (see $M$$M$MM vector in Figure 1.1B), since slightly more magnets are pointing along that direction.
每個質子可以被視為一個小磁鐵。當這些小磁鐵置於沒有任何顯著外部磁場的環境中（例如，在 MRI 掃描儀外），它們將隨機排列（圖 1.1A），因此它們的淨強度，稱為磁化，為 0。另一方面，當這些磁鐵置於強外部磁場中（例如，在 MRI 掃描儀內），它們的排列不再是隨機的。稍微多於一半的磁鐵將會與外部磁場平行，而稍微少於一半的磁鐵將會與外部磁場反平行（圖 1.1B）。因此，質子磁鐵的淨強度不再是 0，而是與外部磁場平行（見圖 1.1B 中的 $M$$M$MM 向量），因為稍微多一些的磁鐵指向該方向。

The magnitude of the net strength, which ultimately determines the intensity of our MRI signal, is given by:
淨強度的大小，最終決定我們 MRI 信號的強度，表達為：

Figure 1.1 Illustration of the formation of MRI signal. (A) When the water protons are placed in an environment without a significant magnetic field, they are oriented randomly, thus the net signal (referred to as magnetization, $M$$M$MM ) is zero. (B) When the water protons are placed in a strong magnetic field, they are aligned along the axis of the external field, ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0}, with a slightly larger fraction parallel to the field (relative to anti-parallel). Thus, the net magnetization is along the $B_0$$B_0$B_(0)B_{0} direction. This net magnetization is what an MRI system measures when performing a scan.
圖 1.1 MRI 信號形成的示意圖。 (A) 當水中的質子置於沒有顯著磁場的環境中時，它們隨機取向，因此淨信號（稱為磁化， $M$$M$MM ）為零。 (B) 當水中的質子置於強磁場中時，它們沿著外部磁場的軸對齊， ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0} ，其中有稍微更大的一部分平行於磁場（相對於反平行）。因此，淨磁化沿著 $B_0$$B_0$B_(0)B_{0} 方向。這個淨磁化是 MRI 系統在進行掃描時所測量的。
where $\rho$$\rho$rho\rho is the proton density (PD) of the issue, ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0} is the strength of the external field, T is the temperature in Kelvin, and C is a parameter related to several physics constants. While mathematically simple, several observations can be made from Equation [1]. First, one can appreciate that the MRI signal is greater at a higher field strength, ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0}. Second, proton density, i.e., the amount of water in the tissue, plays a major role in MRI signal intensity. In the brain, CSF contains the highest water density (about 1 g water $/\mathrm{ml}$$/\mathrm{ml}$//ml/ \mathrm{ml} ) while the gray ( 0.89 g water $/\mathrm{ml}$$/\mathrm{ml}$//ml/ \mathrm{ml} ) and white matter ( 0.73 g water $/\mathrm{ml}$$/\mathrm{ml}$//ml/ \mathrm{ml} ) contain less (Herscovitch and Raichle, 1985). Therefore, in a proton-density MRI, CSF is expected to be brighter than the gray matter, which is in turn brighter than the white matter. Third, the lower the temperature, the greater the MRI signal. Unfortunately, this is not something that one can easily exploit under in vivo conditions.
其中 $\rho$$\rho$rho\rho 是組織的質子密度 (PD)， ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0} 是外部場的強度，T 是以開爾文為單位的溫度，C 是與幾個物理常數相關的參數。雖然在數學上簡單，但從方程式 [1] 中可以得出幾個觀察結果。首先，可以欣賞到在較高的場強下，MRI 信號更強， ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0} 。其次，質子密度，即組織中水的含量，在 MRI 信號強度中扮演著重要角色。在大腦中，腦脊髓液 (CSF) 含有最高的水密度（約 1 克水 $/\mathrm{ml}$$/\mathrm{ml}$//ml/ \mathrm{ml} ），而灰質（0.89 克水 $/\mathrm{ml}$$/\mathrm{ml}$//ml/ \mathrm{ml} ）和白質（0.73 克水 $/\mathrm{ml}$$/\mathrm{ml}$//ml/ \mathrm{ml} ）則含有較少的水（Herscovitch 和 Raichle，1985）。因此，在質子密度 MRI 中，腦脊髓液預期會比灰質更亮，而灰質又會比白質更亮。第三，溫度越低，MRI 信號越強。不幸的是，這在體內條件下並不是容易利用的。

While the application of the external ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0} field can induce a net magnetic field in the tissue, this field along the $B_0$$B_0$B_(0)B_{0} direction cannot be detected. Therefore, a second external field, referred to as radiofrequency (RF) magnetic field (also known as the $B_1$$B_1$B_(1)B_{1} field) is needed to “excite” the tissue spins. With these steps, one can detect a signal in the receiving coil. However, no spatial information is present yet. The spatial information is encoded by applying a magnetic field gradient such that water protons in different spatial locations experience different magnetic fields and therefore have different oscillation frequency. Once these signals are received and recorded by the coil, an image reconstruction algorithm can then be used to obtain the final MRI images.
雖然外部 ${\mathrm{B}}_0$${\mathrm{B}}_0$B_(0)\mathrm{B}_{0} 場的應用可以在組織中誘導出淨磁場，但沿 $B_0$$B_0$B_(0)B_{0} 方向的磁場無法被檢測到。因此，需要第二個外部場，稱為射頻（RF）磁場（也稱為 $B_1$$B_1$B_(1)B_{1} 場），來“激發”組織自旋。通過這些步驟，可以在接收線圈中檢測到信號。然而，尚未存在空間信息。空間信息是通過施加磁場梯度來編碼的，使得不同空間位置的水質子經歷不同的磁場，因此具有不同的振盪頻率。一旦這些信號被線圈接收和記錄，就可以使用圖像重建算法來獲得最終的 MRI 圖像。

The signal one obtains from the final MR image is further modulated by MR properties of the tissue, in particular T1 and T2. Specifically, the MR signal received is given by:
從最終的 MR 影像獲得的信號進一步受到組織的 MR 特性的調制，特別是 T1 和 T2。具體而言，接收到的 MR 信號為：

where FA is flip angle, TR is repetition time, and TE is echo time. In Equation [2], the MR properties T1 and T2 should be differentiated from imaging parameters, e.g., repetition time (TR) and echo time (TE). The difference is that imaging parameters can be chosen by the researcher, whereas T1 and T2 are intrinsic properties of the tissue that the researcher has no control over. However, by properly choosing TR and TE (and other imaging parameters for more advanced sequences), one can make the overall image intensity weighted in a pre-defined fashion, resulting in different contrasts such as T1 weighted or T2 weighted MRI.
其中 FA 是翻轉角度，TR 是重複時間，TE 是回波時間。在方程式 [2] 中，MR 性質 T1 和 T2 應與成像參數區分開來，例如重複時間 (TR) 和回波時間 (TE)。區別在於，成像參數可以由研究者選擇，而 T1 和 T2 是組織的內在特性，研究者無法控制。然而，通過適當選擇 TR 和 TE（以及其他更高級序列的成像參數），可以使整體圖像強度以預定的方式加權，從而產生不同的對比，例如 T1 加權或 T2 加權 MRI。

When evaluating the quality of a set of MRI data, three criteria can be considered. One is to identify whether the image contains any artifact, which is represented by the appearance of signals in unexpected areas. Figure 1.2 illustrates a brain image in which the fat signal was insufficiently suppressed and appeared inside the brain due to chemical shift between fat and water protons. Many reasons can cause image artifacts, and it is preferable to have the images reviewed by an experienced imaging scientist, especially at the beginning of a project, before proceeding with a large sample
在評估一組 MRI 數據的質量時，可以考慮三個標準。一個是識別圖像是否包含任何伪影，這由意外區域出現信號來表示。圖 1.2 顯示了一個腦部圖像，其中脂肪信號未被充分抑制，並因脂肪和水質子之間的化學位移而出現在腦內。許多原因可以導致圖像伪影，最好由經驗豐富的影像科學家對圖像進行審查，特別是在項目開始時，在進行大樣本之前。

Figure 1.2 An example of brain images with and without artifacts induced by undesired fat signal (arrows).
圖 1.2 一個有和沒有由不必要的脂肪信號（箭頭）引起的伪影的腦部影像示例。
scanning. Another criterion is the spatial signal-to-noise ratio (SNR) of the image, which is a useful index in evaluating the quality of structural MR images. The signal in the spatial SNR calculation can be obtained by drawing a region-of-interest (ROI) in a representative brain region and calculating the mean of all voxels, while the noise can be estimated by defining an ROI outside the brain and calculating the mean or standard deviation of the voxels. For MRI data that have multiple time points, such as functional MRI (fMRI) or arterial-spin-labeling (ASL) MRI, a third criterion, temporal SNR, is often used to assess data quality. In the calculation of temporal SNR, the signal term is defined similar to that for spatial SNR. For the noise term, however, it is defined as the standard deviation across the time points. Temporal SNR can be defined on a voxel-by-voxel basis or on an ROI. The estimation of temporal SNR is preferred in fMRI data assessment because it considers contributions from both thermal and physiological noise, whereas spatial SNR only considers thermal noise. Since physiological noise is known to be a major, if not the predominant, component of the noise source in fMRI, the examination of temporal SNR is more appropriate for determining the reliability of a functional dataset.
掃描。另一個標準是影像的空間信號噪聲比（SNR），這是一個評估結構性磁共振影像質量的有用指標。在空間 SNR 計算中，信號可以通過在代表性腦區劃定感興趣區域（ROI）並計算所有體素的平均值來獲得，而噪聲則可以通過在腦外定義 ROI 並計算體素的平均值或標準差來估算。對於具有多個時間點的 MRI 數據，例如功能性磁共振成像（fMRI）或動脈自旋標記（ASL）MRI，通常使用第三個標準，即時間 SNR，來評估數據質量。在時間 SNR 的計算中，信號項的定義類似於空間 SNR。然而，噪聲項的定義是跨時間點的標準差。時間 SNR 可以在每個體素的基礎上或在 ROI 上定義。在 fMRI 數據評估中，時間 SNR 的估算更受青睞，因為它考慮了來自熱噪聲和生理噪聲的貢獻，而空間 SNR 僅考慮熱噪聲。 由於生理噪聲被認為是功能性磁共振成像中噪聲來源的主要成分（如果不是最主要的話），因此檢查時間信噪比更適合用來確定功能數據集的可靠性。

As mentioned above in Equation [1], higher magnetic field strength usually corresponds to a greater sensitivity. 3 T is therefore preferred over 1.5 T for cognitive aging studies. This advantage has been experimentally demonstrated for virtually all brain MRI pulse sequences (although for cardiac and body MRI, 1.5T is sometimes preferred). As far as the magnitude of the enhancement effect is concerned, it depends on specific pulse sequence and spatial resolution. Going from 1.5T to 3T, a typical SNR increase of $50\mathrm{\%}-100\mathrm{\%}$$50\mathrm{\%}-100\mathrm{\%}$50%-100%50 \%-100 \% is often reported (Willinek and Kuhl, 2006; Bradley, 2008). High-resolution scans tend to manifest a greater gain because in those scans thermal noise is usually the dominant source of noise relative to physiological noise, which scales with signal. Some sequences such as BOLD fMRI and ASL perfusion benefit from additional factors related to
如上所述，在方程[1]中，較高的磁場強度通常對應於更大的敏感度。因此，在認知老化研究中，3 T 比 1.5 T 更受青睞。這一優勢在幾乎所有的腦部 MRI 脈衝序列中都得到了實驗證明（儘管在心臟和全身 MRI 中，有時會偏好 1.5T）。至於增強效應的大小，則取決於特定的脈衝序列和空間解析度。從 1.5T 到 3T，通常報告的典型信噪比增益為 $50\mathrm{\%}-100\mathrm{\%}$$50\mathrm{\%}-100\mathrm{\%}$50%-100%50 \%-100 \% （Willinek 和 Kuhl，2006；Bradley，2008）。高解析度掃描往往表現出更大的增益，因為在這些掃描中，熱噪聲通常是相對於生理噪聲的主要噪聲來源，而生理噪聲則隨信號而變化。一些序列，如 BOLD fMRI 和 ASL 灌注，受益於與之相關的額外因素。
enhanced magnetic susceptibility effects and longer T1 at 3T. Other sequences such as FLAIR benefit less because slower signal recovery at higher field offsets some of the advantages.
增強的磁敏感性效應和在 3T 下更長的 T1。其他序列如 FLAIR 的好處較少，因為在較高的場強下信號恢復較慢，抵消了一些優勢。

7T MRI is also becoming increasingly available in some research institutions. There are about 50 to 60 human 7T systems around the world. However, the use of 7T in cognitive aging studies is still in an early stage. Despite the promise of an increased sensitivity, 7T MRI still suffers from several technical limitations at present. These limitations are due to magnetic field inhomogeneity (thereby resulting in image inhomogeneity), higher power deposition (thus tissue heating becomes an issue, which is usually not a problem when going from 1.5 T to 3 T ), and physiological side effects (i.e., participant may feel dizzy when entering the scanner), aside from its high costs. However, these limitations may be resolved with technical efforts and advances, as shown by highly promising results from the Human Connectome Project.
7T 磁共振成像在一些研究機構中也變得越來越普及。全球大約有 50 到 60 台人類 7T 系統。然而，7T 在認知老化研究中的使用仍處於早期階段。儘管提高靈敏度的前景令人期待，但目前 7T 磁共振成像仍然面臨幾個技術限制。這些限制是由於磁場不均勻性（因此導致影像不均勻性）、更高的功率沉積（因此組織加熱成為一個問題，這在從 1.5 T 到 3 T 的過程中通常不是問題）以及生理副作用（即參與者在進入掃描儀時可能會感到頭暈），此外還有其高昂的成本。然而，這些限制可能會通過技術努力和進步得到解決，正如人類連接組計劃所顯示的非常有前景的結果。

It is recommended that each scan session be less than 60 minutes, as excessive motion is often observed when the subject has been inside the scanner for a long period of time. When that happens, the data collected are of low quality and usually need to be excluded. This is especially likely for elderly participants. For studies that require more than 60 minutes, one should consider allowing the subject to come out of the scanner to take a break before entering again for the remainder of the scans.
建議每次掃描會話少於 60 分鐘，因為當受試者在掃描儀內待的時間過長時，通常會觀察到過度運動。當這種情況發生時，收集到的數據質量較低，通常需要排除。這對於老年參與者尤其可能發生。對於需要超過 60 分鐘的研究，應考慮允許受試者在再次進入掃描儀進行剩餘掃描之前先出來休息。

It is also useful to prioritize the scans such that the pulse sequences that are most important for the study hypothesis are performed first. It is not uncommon that a subject would decide to abort the scan session or show excessive motion after staying in the scanner for a while. Thus, arranging the scan order based on priority can ensure the successful data collection of the most relevant sequences. Some investigators also found it helpful to place the functional and physiological scans at the beginning of the session when the subject is most alert and attentive. The structural scans can usually be performed even when the subject is asleep.
優先安排掃描是有用的，這樣對於研究假設最重要的脈衝序列可以優先進行。受試者在掃描儀中待了一段時間後，決定中止掃描會話或出現過度運動的情況並不罕見。因此，根據優先順序安排掃描順序可以確保成功收集最相關序列的數據。一些研究者還發現，將功能性和生理性掃描放在會話開始時進行是有幫助的，因為此時受試者最為清醒和專注。結構性掃描通常即使在受試者睡著的情況下也可以進行。

When calculating the total scan session duration, it should be remembered that the scan duration displayed on the scanner console often underestimates the real-life scan time, as the preparation time necessary at the beginning of every scan is usually not included in the displayed time. The preparation time is inherent to every MRI pulse sequence and usually includes $B_0$$B_0$B_(0)B_{0} shimming, RF center frequency determination, RF power optimization, and dummy scans to allow the magnetization to approach a steady state. Collectively, the actual data collection time for a 60-minute session may be around 45 minutes. It is useful to keep this in mind during planning. A small number of pilot scans should be performed before finalizing the protocol and obtaining a more accurate estimation of the session duration.
在計算總掃描會話持續時間時，應該記住掃描儀控制台上顯示的掃描持續時間通常低估了實際的掃描時間，因為每次掃描開始時所需的準備時間通常不包括在顯示的時間內。準備時間是每個 MRI 脈衝序列固有的，通常包括 $B_0$$B_0$B_(0)B_{0} 的調整、射頻中心頻率確定、射頻功率優化和虛擬掃描，以使磁化接近穩定狀態。總體而言，60 分鐘會話的實際數據收集時間可能約為 45 分鐘。在計劃過程中記住這一點是有用的。在最終確定協議並獲得更準確的會話持續時間估算之前，應進行少量的試點掃描。

Table 1.1 provides a list of MRI techniques commonly used in cerebral aging studies. They measure various aspects of the brain structure and function. Later chapters will
表 1.1 提供了在腦部老化研究中常用的 MRI 技術列表。它們測量大腦結構和功能的各個方面。後面的章節將

Table 1.1 List of MRI techniques and associated imaging parameters at 3T.
表 1.1 3T MRI 技術及相關影像參數列表。