专长领域功能性磁振造影（FMRI）课件.ppt

Functional Imaging Laboratory (FIL) * Some Terminologies x: input features Eg. x =[ weight, height, waist line, chest circumference, age], x=[Brain volume in different regions (region1, region2….region n)] Dimensionality : number of features in x Eg. x =[ weight, height, age] The dimension is 3 f(x,θ) x y y : output, it can be a continuous value or a discrete value (classes). Eg. continuouse values: y =[age(12)], Eg, classes: y=[‘male’], y=[‘female’] F(x, θ) : The Model which estimates the output y form the given input x features with model parameters θ. Examples of simple model f(x, θ) Gaussian Model 2 parameters (mean and standard deviation) Linear Model 2 parameters (gradient and offset) x1(feature1) X2(feature2) Two sets of data x1(feature1) X2(feature2) Class 1 Class 2 x1(feature1) X2(feature2) Gaussian Model Class 1 Class 2 x1(feature1) X2(feature2) Class 1 Class 2 Linear Discriminative function x1(feature1) X2(feature2) But there are many possible separating lines x1(feature1) X2(feature2) Maximizing the margin γ Support Vector Machine Separating line x1(feature1) X2(feature2) Maximizing the margin γ Support Vector Machine Suppor vector Suppor vectors Separating line 12 Queen Square, St. John’s House About 60 Researchers around the globe Different Groups MR Physics Analysis Methods (where I belong to) Language Emotion. Etc… Primary Supervisor Dr.John Ashburner 職稱??Honorary Senior Lecturer 單位??The Wellcome Department of Imaging Neuroscience 專長領域??功能性磁振造影（fMRI）、神經解剖運算模型(computational Neuroamatomy) Quoted from 第一屆台灣地區認知神經科學暑期學校 Ashburner 博士目前任職於英國的 FIL，是在腦訊號處理方法學及物理學方面的主要研究員，主要專長是計算神經解剖學。由於我們由腦造影工具所獲得的影像需要倚賴一套適當的運算模型，來調整個別受試者的腦部不同形狀大小以獲得具代表性的空間訊號，才能作進一步的訊號比較，而 FIL 能夠在腦訊號處理科學界不但是開路先驅更於現在位居龍頭，Ashburner 博士所奠定的基礎功不可沒。 Secondary Supervisor: Professor Karl Friston Position: Director of Functional Imaging Laboratory Founders of SPM software package Fellow of the Royal Society Quoted from The Royal Society Professor Friston is Professor of Neurosci