使用全腦神經纖維束分析方法探討擴散影像以不同的方式採樣造成之影響

Effects of different sampling schemes of diffusion imaging on tract-based analysis

楊曜彰(Yao-Chang Yang)

指導教授 ： 曾文毅

國立臺灣大學/醫學院/醫療器材與醫學影像研究所/碩士(2018年)

https://doi.org/10.6342/NTU201900344

摘要

研究目的: 磁振造影(Magnetic Resonance Imaging, MRI)用於腦部掃描已行之有年，磁振造影的機器與方法隨著科技的進步，也不停地推陳出新，但是這也造成了一個問題，經由不同機器與不同成像方式得到的成果卻不能拿來一起比較，這使得研究結果的再現性與比較性容易受到很大的質疑。近年來多殼層擴散造影(Multi-shell diffusion imaging, MSI)受到很大的重視，因為它的模組化，B值(b-value)與取樣點數(sampling points)的可調性，可以彈性的選取所需要的B值與點數來得到想要的研究結果，這代表只要選取相近B值與點數的MSI可以大大降低比較性的質疑，但是MSI有一個很大的缺點，它所需的掃描時間由於取樣點數的過多而較長，對於受試者是一個很大的負擔，所以我們想要藉由減少B值與取樣點數來達到可以利用於臨床研究使用的程度，也可以藉此探討B值與取樣點數對研究結果所造成的影響。
研究方法: 35位健康受測者的MSI掃描資料是由麻省總醫院(Massachusetts General Hospital, MGH)的雲端資料庫提供，MSI是在3T磁振造影儀上掃描並由四個殼層B值1000, 3000, 5000, 10000與相對應的取樣點數64, 64, 128, 256共512點所構成，平均一人掃描時間為89分鐘，由於我們實驗室有做影像品質篩選(Quality assessment, QA)，我們篩選掉了其中四位與其他人品質差距過大的受試者，故剩餘31位，再來我們將原始的MSI資料共512點，經過演算法做等比降低點數的取樣方式，分成各種組合，這些組合透過全腦基於神經束之自動化分析(Tract-based automatic analysis, TBAA)來擷取全腦主要76條神經纖維束之平均表觀傳播系數(Mean Apparent Propagator MRI index, MAP-MRI index)與擴散張量系數(Diffusion tensor index)，透過這些係數來評估臨床使用上的可行性與B值、取樣點數的改變所造成的影響，由於需要客觀的比較我們選了三個MSI的組合做為標準，總共512點、最大B值為10000的組合為第一個標準，用以當作MAP-MRI index標準值；總共256點、最大B值為5000的組合作為第二個標準，用以當作臨床實驗的標準值；總共64點、B值為1000的組合做為第三個標準，用以評估擴散張量系數。
研究結果與討論: 我們發現當我們將MSI點數降到B值10000，128點與標準一做比較時，發現MAP-MRI係數並沒有太大的差距，都低於10個百分比，這代表者即使我們大大的降低了MSI的取樣點數，還是能得到可以接受的MAP-MRI係數。在擴散張量係數方面，我們發現我們拿標準三與其他組合做比較時，其百分比差都大於20甚至更多，我們從中得知如果直接拿整個MSI下去取樣擴散張量係數的話，所得到的結果跟標準差距很大，最好還是抽取MSI中B值為1000的部分作為比較，才能得到較為正確的擴散張量係數。臨床實驗用標準的部分，我們取B值5000，點數為64點的組合與標準二做比較發現MAP-MRI係數差距都低於10個百分點，而且這64點的組合，掃描時間僅需11分鐘，接近一般臨床使用上的掃描時間。
我們的實驗主要探討了MSI的可塑性與它在臨床使用上的可行性，我們成功地降低了取樣點數與B值卻又能得到不錯的MAP-MRI係數表現，也得到了能夠使用在臨床上的MSI組合，在未來的研究中，我們將會確認這些組合能夠穩定的發會在它所適用的領域中。

關鍵字

多殼層擴散造影 ； B值 ； 取樣點數 ； 平均表觀傳播系數 ； 擴散張量係數 ； 全腦基於神經束之自動化分析

並列摘要

Introduction: Magnetic Resonance Imaging (MRI) has been used for brain scanning for many years. The machines and methods of magnetic resonance imaging continue to evolve with the advancement of technology. However, this also creates a problem. The results obtained by different machines and different imaging methods cannot be compared together, which makes the reproducibility and comparativeness of the research results be questioned. In recent years, multi-shell diffusion imaging (MSI) has received great attention because of its modularity and the adjustment of sampling points, b-value, which make it easily to obtain the desired research results. This means as long as the selection of MSI with similar b-values and sampling points can reduce the comparative doubts.
However, MSI has a big disadvantage, the scan time required for it is too long due to the large number of sampling points, which is a big burden for the subject. Therefore, we want to reduce the b-value and the number of sampling points to make it be used in clinical research, and also explore the effects of b-value and sampling points on the research results.
Material and method: 35 healthy subjects’ MSI data were provided by the database of the Massachusetts General Hospital (MGH). The MSI is scanned on a 3T CONNECTOM scanner and consists of four shell b-values of 1000, 3000, 5000, 10000 and corresponding sampling points of 64, 64, 128, 256 points. The average one-person scan time is 89 minutes. After we did quality assessment (QA), we screened out four of the subjects with excessive quality differences from others, and the remaining 31 subjects. Then we used the MSI data with total 512 sampling points, through the algorithm reducing the sampling points, divided into various sets. We used the Tract-based automatic analysis (TBAA) to measure the Mean Apparent Propagator MRI index (MAP-MRI index) and diffusion tensor index of the 76 major tracts in the whole brain. Through these index, the feasibility of clinical use and the effects of changes in b-values and sampling points are evaluated.
Due to the need for comparison, we selected three MSI sets as reference standards. A set with total 512 sampling points and maximum b-value 10000 was the first standard to be used as the MAP-MRI index standard; A set with total 256 sampling points and maximum b-value 5000 was the second standard for clinical trials; A set with total 64 sampling points and maximum b-value 1000 was the third standard to be used as the diffusion tensor index standard.
Results and Discussions: We found that when we reduced the MSI sampling points to 128 points compared with the first standard, the MAP-MRI index did not have much difference, which was less than 10 percentage points. This represented even if we greatly reduced the number of sampling points of the MSI we can still obtain an acceptable MAP-MRI index.
In terms of the diffusion tensor index, we found that when we compared the third standard with other sets, the percentage difference is greater than 20 or more. We know that if we take the entire MSI to sample the diffusion tensor coefficient, the obtained results are very different from the standard. It is better to extract the portion of the MSI with b-value 1000 as a comparison to obtain a more accurate diffusion tensor index.
For the standard of clinical trials, we compared the set with b-value 5000 and 64 sampling points with the second standard and found that the MAP-MRI index difference was less than 10 percentage points, and the scan time required only 11 Minutes, which was close to the scan time in general clinical use.
Our research mainly found the adjustability of MSI and its feasibility in clinical use. We successfully reduced the number of sampling points and b-value, but also got a good performance. In future research, we will confirm these sets can be stable in their appropriate area.

並列關鍵字

Multi-shell diffusion imaging ； Mean apparent propagator MRI index ； Diffusion tensor index ； sampling points ； b-value ； Tract-based automatic analysis

參考文獻

Chen, Y. J., Lo, Y. C., Hsu, Y. C., Fan, C. C., Hwang, T. J., Liu, C. M., . . . Tseng, W. Y. (2015). Automatic whole brain tract-based analysis using predefined tracts in a diffusion spectrum imaging template and an accurate registration strategy. Hum Brain Mapp, 36(9), 3441-3458. doi:10.1002/hbm.22854

Ozarslan, E., Koay, C. G., Shepherd, T. M., Komlosh, M. E., Irfanoglu, M. O., Pierpaoli, C., & Basser, P. J. (2013). Mean apparent propagator (MAP) MRI: a novel diffusion imaging method for mapping tissue microstructure. Neuroimage, 78, 16-32. doi:10.1016/j.neuroimage.2013.04.016

Wedeen, V. J., Wang, R. P., Schmahmann, J. D., Benner, T., Tseng, W. Y., Dai, G., . . . de Crespigny, A. J. (2008). Diffusion spectrum magnetic resonance imaging (DSI) tractography of crossing fibers. Neuroimage, 41(4), 1267-1277. doi:10.1016/j.neuroimage.2008.03.036

Alexander, A. L., Lee, J. E., Lazar, M., & Field, A. S. (2007). Diffusion tensor imaging of the brain. Neurotherapeutics, 4(3), 316-329. doi:10.1016/j.nurt.2007.05.011

Ashburner, J., & Friston, K. J. (2000). Voxel-based morphometry--the methods. Neuroimage, 11(6 Pt 1), 805-821. doi:10.1006/nimg.2000.0582