本研究主要為最佳傳輸理論(optimal transport theory) 與資料擴增演算法(data augmentation algorithm) 運用於腦部腫瘤病灶分析；人體的惡性腫瘤為目前全球的10 大死因之一，若能在早期檢測出疾病的存在，可以有效幫助治療，在醫學研究與臨床中，腦腫瘤的主要診斷方式是透過醫學影像，腫瘤的檢測透過核磁共振成像或是電腦斷層掃描以及超聲波圖像。
本文以3D Unet 作為機器學習的演算法，並在其中應用了最佳傳輸以及資料擴增，希望在有限的臨床資料裡能夠獲得更高的病灶判斷準確度，以協助相關領域的研究人員與從業人員將影像辨識技術應用於臨床病症。

[1] Wang,G.,etal,Automaticbraintumorsegmentationusingcascaded
anisotropic convolutionalneuralnetworks. Brainlesion:Glioma,Multi-
ple Sclerosis,StrokeandTraumaticBrainInjuries, 178 − 190(2018)
[2] Parihar,A.S.Astudyonbraintumorsegmentationusingconvolution
neural network.2017InternationalConferenceonInventiveComputing
and Informatics(ICICI) (2017).
[3] I. R.Haque,J.Neubert.Deeplearningapproachestobiomedicalimage
segmentation.https://doi.org/10.1016/j.imu.2020.100297/(2020)
[4] J.Kleesiek ,G.Urbanetal,DeepMRIbrainextraction:A3Dconvo-
lutional neuralnetworkforskullstripping, NeuroImage, Volume 129,
April 2016, 460 − 469(2016).
[5] Paul,T.BrainMagneticResonanceImaging(MRI)asaPotential
BiomarkerforParkinson’sDisease(PD) (2017).
[6] Louis, D.N,etal.The2016worldhealthorganizationclassi
cation oftumorsofthecentralnervoussystem:asummary. ActaNeu-
ropathologica 131(6), 803 − 820(2016)
[7] Menze, B.H,etal.Themultimodalbraintumorimagesegmentation
benchmark(BRATS). IEEE TransactionsonMedicalImaging 34(10),
1993 − 2024(2015)
[8] Bakas, S.,etal.AdvancingthecancergenomeatlasgliomaMRIcol-
lections withexpertsegmentationlabelsandradiomicfeatures. Nature
Scientic Data p. 170117(2017)
[9] G.Wang,etal.AutomaticBrainTumorSegmentationusingConvolu-
tional NeuralNetworkswithTest-TimeAugmentation. https://arxiv.
org/pdf/1810.07884.pdf (2018)
[10] G.Wang,etal.Automaticbraintumorsegmentationusingcascaded
anisotropic convolutionalneuralnetworks.Brainlesion:Glioma,Mul-
tiple Sclerosis,StrokeandTraumaticBrainInjuries 178 − 190(2018)
[11] Andriy,M.3DMRIbraintumorsegmentationusingautoencoderregu-
larization. https://arxiv.org/pdf/1810.11654.pdf(2018)
[12] Havaei,M.,etal.Braintumorsegmentationwithdeepneuralnetworks.
MedicalImageAnalysis 35, 18 − 31(2016)
[13] Kamnitsas, K.,etal.Efficientmulti-scale3DCNNwithfullyconnected
CRF foraccuratebrainlesionsegmentation. MedicalImageAnalysis
36, 61 − 78(2017)
[14] Long, J.,etal.Fullyconvolutionalnetworksforsemanticsegmentation.
IEEE ConferenceonComputerVisionandPatternRecognition 3431 −
3440 (2015)
[15] Ronneberger,O.,etal.U-Net:Convolutionalnetworksforbiomedical
image segmentation. International ConferenceonMedicalImageCom-
puting andComputer-AssistedIntervention 234 − 241(2015)
[16] Bonnotte, N.FromKnothe’s rearrangementtoBrenier’s optimaltrans-
portmap. SIAM J.Math.Anal 45(1), 64–87(2013)https://doi.org/
10.1137/120874850
[17] Kantorovich,L.V.Onaproblemofmonge. UspekhiMat.Nauk 3, 225−
226(1948).
[18] Garg, V.,etal.AdvancesinNeuralInformationProcessingSystems 32,
8014 − 8025 Curran Associates,Inc(2019). https://arxiv.org/abs/
1905.12158
[19] Su, Z.,etal.Optimalmasstransportforshapematchingandcompari-
son. IEEETrans.PatternAnal.Mach.Intell. 37(11), 2246–2259(2015).
https://doi.org/10.1109/TPAMI.2015.2408346
[20] Lei, N.,etal.Ageometricviewofoptimaltransportationandgenerative
model.Comput.AidedGeom.Des. 68,1–21(2019). https://doi.org/
10.1016/j.cagd.2018.10.005
[21] Marcinkiewicz M.,etal.SegmentingbraintumorsfromMRIusingcas-
caded multi-modalU-Nets.Glioma,MultipleSclerosis,Strokeand
TraumaticBrainInjuries-4thInternationalWorkshop.RevisedSe-
lected Papers,PartII,vol.11384ofLectureNotesinComputerScience
,13–24(2018).
[22] Jakub, N.,etal.DataAugmentationforBrain-TumorSegmentation:A
Review. https://doi.org/10.3389/fncom.2019.00083
[23] Marco,M. C.,etal.Whatisthebestdataaugmentationfor3Dbrain
tumor segmentation? https://arxiv.org/pdf/2010.13372.pdf
[24] PatriceY.S.,etal.BestPracticesforConvolutionalNeuralNetworks
Applied toVisualDocumentAnalysis. (2003)
[25] Su, K.,etal.Volumepreservingmeshparameterizationbasedonoptimal
mass transportation.Comput.AidedDes. 82, 42–56(2017). https://
doi.org/10.1016/j.cad.2016.05.020
[26] Gu, X.,etal.VariationalprinciplesforMinkowskitypeproblems,dis-
crete optimaltransport,anddiscreteMonge–Ampèreequations.Asian
J. Math. 20(2), 383–398(2016). https://doi.org/10.4310/AJM.2016.
v20.n2.a7
[27] Brenier, Y.Polarfactorizationandmonotonerearrangementofvector-
valuedfunctions.Commun.PureAppl.Math. 44(4), 375–417(1991).
https://doi.org/10.1002/cpa.3160440402
[28] Yueh,M.H.,Li,T.,Lin,W.W.,Yau,S.T.Anovelalgorithmfor
volume-preservingparameterizationsof3-manifolds.SIAMJ.Imag.Sci.
102(2), 1071–1098(2019). https://doi.org/10.1137/18M1201184
[29] W.W.Lin, C.Juang,M.H.Yueh,T.M.Huang,T.Li,S.Wang&
S.T.Yau.3Dbraintumorsegmentationusingatwo-stageopti-
mal masstransportalgorithm. https://www.nature.com/articles/
s41598-021-94071-1
[30] Create 3-DU-Netlayersforsemanticsegmentationofvolumetricimages:
MATLABunet3dLayers. https://www.mathworks.com/help/vision/
ref/unet3dlayers.html.