利用體素內非均向性運動技術於腦部中風偵測系統之開發

Development of Stroke Detection System Using Intravoxel Incoherent Motion Method

林庭偉(Ting-Wei Lin)

指導教授 ： 蘇振隆

中原大學/工學院/生物醫學工程研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000689

摘要

摘要
急性腦中風(acute ischemic stroke, AIS)的病患，必須在發病後短時間內，做出診斷，選擇治療的方式，才能得到好的癒後。但由於少數病患無法施打現影劑，做進一步的評估，而延遲治療時間，本研究的目的是在不施打對比劑的情形下，發展一套輔助軟系統，給予醫師更多評估的資訊。
本研究使用飛利浦1.5特斯拉(Tesla)之磁振造影儀，做腦部磁振造影(Magnetic Resonance Image, MRI)掃描，收取人體腦部MRI影像及16個不同b值之擴散加權影像(diffusion weighted imaging, DWI)。利用Matlab軟體，產生貝葉斯模型之程式，擬合(fitting)體素內非均向性運動(intravoxel incoherent motion, IVIM)雙指數模型(bi-exponential)，進而得到D(diffusion coefficient)擴散係數、F(perfusion fraction)灌注分數、D*(pseudo-diffusion coefficient)非均向性流體擴散係數及ADC(apparent diffusion coefficient)表觀擴散係數等參數。最後並用臨床影像評估本系統。
運用貝葉斯模型發展好的系統包括DICOM影像讀取、圈選相同組織、產生迭代運算初始值、計算直方圖訊號分佈、貝葉斯計算擬合IVIM雙指數模型。透過隨機挑選5名中風之病患做參數評估，結果發現D*的變異性太大、ADC值與D值相比也並不是十分準確。所以本研究將D值與F值作為此次分析參數。利用32名缺血性腦中風之病患及5名腦部結構正常病患影像中選擇之92處，透過不同之D值與F值的參數數據，計算組織影像強度比(tissue image intensity ratio)，經與醫師討論後設定中風之閾值。系統於選定以F值及D值均為0.98當作中風閾值時表現最佳；其靈敏度、特異性、準確度、Kappa值分別為0.889、0.85、0.888、0.677及0.919、0.7、0.87、0.618。
本研究結果發現利用貝葉斯法擬合IVIM雙指數模型於中風之偵測是可行的，貝葉斯模型沒有使用者定義參數的不確定性，也解決了LM法逐個像素擬合時，產生大量雜訊的情形。但本系統對於左、右腦皆患有血管異常或腦組織病變者較有限制，期望未來針對數據上顯示有血管自行灌注之患者做進一步的分析與驗證。

關鍵字

急性缺血性腦中風 ； 擴散加權影像 ； 體素內非均向性運動 ； 貝葉斯模型

並列摘要

Abstract
Patients with acute stroke (acute stroke, AIS) must make a diagnosis and choose the treatment method shortly after the onset of the disease in order to get a good recovery. However, due to the inability of a small number of patients to give a contrast agent for further evaluation and delay treatment time, the purpose of this study is to develop a computer aids diagnosis (CAD) system to give physicians more information for evaluation without the use of contrast agents.

This study used a Philips 1.5 Tesla Magnetic Resonance Image (MRI) system to scan the brain, 16 difference b values for diffusion-weighted images (DWI) are collected. Matlab software is used to generate a Bayesian model program, and the signals obtained by different b values are fitted into the intra-voxel incoherent motion (IVIM) bi-exponential model, and then diffusion coefficient (D), perfusion fraction (F), pseudo-diffusion coefficient (D*), and apparent diffusion coefficient (ADC) can be determined. Finally, the system is evaluated by clinical imaging.

Systems developed include DICOM image reading, circle the same tissue, produce iterative operation initial values, calculate histogram signal distribution, and Bayesian calculate fit IVIM bi-exponential models. By randomly selecting 5 stroke patients for parameter evaluation, it was found that the variability of D* was too large and that the ADC value was not very accurate compared to the D value. Therefore, this study used the D and F values as parameters for further study. Using 92 ROI of the 32 ischemic stroke patients and 5 patients with normal brain structure, the tissue image intensity ratio was calculated using parameter data from different D and F values, and the stroke threshold was set after discussion with a physician. The system performed best when it selected 0.98 as a threshold with both F and D value, and its sensitivity, specificity, accuracy, Kappa values were 0.889, 0.85, 0.888, 0.677, and 0.919, 0.7, 0.87, 0.618, respectively.

The results of this study show that it is feasible to detect stroke by using Bayesian method fitting IVIM bi-exponential model. This model does not have the uncertainty of user-defined parameters, and also solves the situation of generating a lot of noise when LM method fits pixel by pixel. However, the system for the left and right brain are suffering from vascular abnormalities or cerebral tissue lesions are more limited, it is expected that in the future for the data show that there are vascular self-perfusion patients to do further analysis and verification.

並列關鍵字

acute ischemic stroke (AIS) ； diffuse weighted imaging (DWI) ； intra-voxel incoherent motion (IVIM) ； Bayesian model.

參考文獻

[1] 衛生福利部.[online].2015.Available from: https://www.mohw.gov.tw/cp-16-33598-1.html (最後修訂2015/6/19，2020/6/25資料擷取)

[2] 衛生福利部.[online].2016.Available from: https://www.mohw.gov.tw/cp-3795-41794-1.html (最後修訂2016/8/2，2020/6/25資料擷取)

[3] 台灣腦中風學會: 靜脈血栓溶解劑治療急性缺血性腦中風指引 2013

[4] J Bamford, P Sandercock, M Dennis, C Warlow, J.Burn, "Classification and natural history of clinically identifiable subtypes of cerebral infarction." The Lancet, 337.8756: 1521-6, 1991.

[5] LM Allen, AN Hasso, J Handwerker, H Farid, "Sequence-specific MR imaging findings that are useful in dating ischemic stroke." Radiographics, 32.5: 1285-97, 2012.