慢性肝病常會進行到不可逆的肝硬化階段，但是在肝纖維化的早期，其病勢是可逆的。
雖然侵入式診斷是目前評斷肝纖維化的黃金標準，除了成本較為高昂且會造成病人的
痛苦且帶有些微機率的後遺症。所以我們希望能建立一套簡單但可靠的非侵入式評分
方法，並期望能判別患者肝纖維化的程度。主要影像是使用小鼠的MR 影像，利用
MR 影像的T1 和T1ct (Primovist) 各自的成像差異加上K-means 法把肝影像分割出來。我們架構的模型主要使用Support vector machine (SVM) 去進行分類。透過觀察,我們使用simple connected domain 的機率當作其中的一個特徵，並用孔隙率當作第二個特徵。由於SVM 本身是Binary 分類器，所以我們選用One-Against-One 策略，使用多個Binary 分類器去達到我們想要的分類效果。由於資料組數較少，因此我們選用K-Fold 交錯驗證法來增強結果的可信度。單獨使用孔隙率當作特徵所得到的模型準確率大約有80%，根據ROC 曲線分析，可以得出其AUROC 分數大約為0.7 ∼ 0.8 之間，但如果加上simple connected domain 作為第二特徵，訓練出來的模型準確率可達 90%，AUROC 分數可達將近0.9，代表其判斷結果是可信的。

Chronic liver disease often progresses to the irreversible stage of cirrhosis, but in the early
stage of liver fibrosis, its disease is reversible. Although the invasive diagnosis is currently
the gold standard for judging liver fibrosis, in addition to the high cost and the painful
sequelae of the patient with some micro-probability. Therefore, we hope to establish a
simple but reliable non-invasive scoring method and hope to determine the degree of liver
fibrosis in patients. The main image is the MR image of the mouse. The liver image is
segmented using the respective imaging differences of T1 and T1ct (Primovist) of MRI
plus the K-means method. The model of our architecture mainly uses a support vector
machine (SVM) to classify. Through our observation, we use the degree of the simple
connected domain as one of the features, and the porosity as the second feature. Since
SVM is a binary classifier, we choose the one-against-one mode and use multiple binary
classifiers to achieve the classification effect we want. Due to the small number of data
sets, we choose K-Fold cross-validation to enhance the credibility of the results. The
accuracy of the model obtained by using porosity alone as a feature is only about 80%.
According to the ROC curve analysis, it can be concluded that its AUROC score is about
0.7∼0.8, but if the simple connected domain is added as the second feature, the accuracy
of the trained model can reach 90%, and the AUROC score can reach close to 0.9, which means that the judgment result is credible.

[1] J. Axnick. β1 integrin-dependent mechanotransduction induces angiocrine signaling
required to promote liver growth and survival. PhD thesis, 2019.
[2] J.W. Chen. Exploring effects of optimizer selection and their hyperparameter tuning
on performance of deep neural networks for image recognition. Master’s thesis,
National Central University, 2019.
[3] U.R. Kamalov A.V. Kvostikov, A.S. Krylov. Ultrasound image texture analysis for
liver fibrosis stage diagnostics. Programming and Computer Software, 41:273–278,
2015.
[4] G.N Hounsfield. Computerized transverse axial scanning (tomography): Part 1.
description of system. The British Journal of Radiology, 46:1016–1022, 1973.
[5] A. Kumar, D. Welti, and R.P Ernst. Imaging of macroscopic objects by NMR Fourier
zeugmatography. Naturwissenschaften, 62:34–34, 1975.
[6] S.C. Faria, K. Ganesan, I. Mwangi, M. Shiehmorteza, B. Viamonte, S. Mazhar,
M. Peterson, Y. Kono, C. Santillan, G. Casola, et al. MR imaging of liver fibrosis:
current state of the art. Radiographics, 29:1615–1635, 2009.
[7] L. Petitclerc, G. Gilbert, B.N. Nguyen, and A. Tang. Liver fibrosis quantification by
magnetic resonance imaging. Topics in Magnetic Resonance Imaging, 26:229, 2017.
[8] K. Sandrasegaran, F.M. Akisik, C. Lin, B. Tahir, J. Rajan, R. Saxena, and A.M.
Aisen. Value of diffusion-weighted MRI for assessing liver fibrosis and cirrhosis.
American Journal of Roentgenology, 193:1556–1560, 2009.
[9] T.L. Harada, K. Saito, Y. Araki, J. Matsubayashi, T. Nagao, K. Sugimoto, and
K. Tokuuye. Prediction of high-stage liver fibrosis using adc value on diffusionweighted
imaging and quantitative enhancement ratio at the hepatobiliary phase of
gd-eob-dtpa–enhanced MRI at 1.5 t. Acta Radiologica, 59:509–516, 2018.
[10] C.H. Thng and D.J. .and Koh D.M. San Koh, T.and Collins. Perfusion magnetic
resonance imaging of the liver. World Journal of Gastroenterology: WJG, 16:1598,
2010.
[11] M. Yin, K.J. Glaser, J.A. Talwalkar, J. Chen, A. Manduca, and R.L. Ehman. Hepatic
MR elastography: clinical performance in a series of 1377 consecutive examinations.
Radiology, 278:114–124, 2016.
[12] H. Morisaka, U. Motosugi, S. Ichikawa, T. Nakazawa, T. Kondo, S. Funayama,
M. Matsuda, T. Ichikawa, and H. Onishi. Magnetic resonance elastography is as
accurate as liver biopsy for liver fibrosis staging. Journal of Magnetic Resonance
Imaging, 47:1268–1275, 2018.
[13] T. Yokoo, T. Wolfson, K. Iwaisako, M.R. Peterson, H. Mani, Z. Goodman,
C. Changchien, M.S. Middleton, A.C. Gamst, S.M. Mazhar, et al. Evaluation of
liver fibrosis using texture analysis on combined-contrast-enhanced magnetic resonance
images at 3.0 t. BioMed research international, 2015.
[14] Y. Yu, J. Wang, C.W. Ng, Y. Ma, S. Mo, E.L.S. Fong, J. Xing, Z. Song, Y. Xie,
K. Si, et al. Deep learning enables automated scoring of liver fibrosis stages. Scientific
reports, 8:1–10, 2018.
[15] T.H. Kim, J.E. Kim, J.H. Ryu, and C.W. Jeong. Development of liver surface
nodularity quantification program and its clinical application in nonalcoholic fatty
liver disease. Scientific reportz, 9:1–10, 2019.
[16] W.G. Bradley D.D. Stark. Magnetic resonance imaging; the cv mosby company: St.
Louis, Washington DC, Toronto, pages 161–181, 1988.
[17] A.D. McLachlan A. Carrington. Introduction to magnetic resonance: with applications
to chemistry and chemical physics. 1967.
[18] Y.T. Juan. Three-dimensional geometry reconstruction of mouse liver from MR
images using k-means method with confusion component removing. Master’s thesis,
National Central University, 2019.
[19] M. Singh, P. Patel, D. Khosla, and T. Kim. Segmentation of functional MRI by
k-means clustering. IEEE Transactions on Nuclear Science, 43:2030–2036, 1996.
[20] H.P. Ng, S.H. Ong, K.W.C. Foong, P.S. Goh, and W.L. Nowinski. Medical image
segmentation using k-means clustering and improved watershed algorithm. In 2006
IEEE southwest symposium on image analysis and interpretation, pages 61–65. IEEE,
2006.
[21] K.S. Sindhushree, T.R. Manjula, and K. Ramesha. Detection and 3d reconstruction
of brain tumor from brain MR images. International Journal of Engineering Research
& Technology (IJERT), 2:528–534, 2013.
[22] R.C. Gonzalez, R.E. Woods, and S.L. Eddins. Digital image processing using MATLAB.
Pearson Education India, 2004.
[23] P. Refaeilzadeh, L. Tang, and H. Liu. Cross-validation. Encyclopedia of database
systems, 5:532–538, 2009.
[24] J.D. Rodriguez, A. Perez, and J.A. Lozano. Sensitivity analysis of k-fold cross validation
in prediction error estimation. IEEE transactions on pattern analysis and
machine intelligence, 32:569–575, 2009.
30