基本脉冲序列与MR图像. Image Characteristics. 上次课主要内容上次课主要内容l层选层选l频率编码频率编码l相位编码相位编码lK空间空间l1D/2D/3D 采集采集. 9/16/20244Slice SelectionGzRFResonantFrequency. . . 1D 采集采集 - Line by Line, Line Scan. Phase Encoding. 9/16/202492D Spatial Encoding – Frequency and Phase EncodingGxPhase encodingPhase ShiftFrequency X-PositionPhase Y-Position. For 2D imagingImaging Equation? k-Space? . k-SpacelFrequency-Encoded SignalsGxtkx0kx0kx0180. k-SpacelPhase-Encoded SignalstGykykx. kykxFor 2D imagingImaging Equation? . kykxFT-1. 3D 采集采集3D与与2D采集的区别采集的区别RF脉冲是非选择性的脉冲是非选择性的选层梯度场选层梯度场 相位相位编码编码3D FT扫描时间增加扫描时间增加(因为因为相位编码步数增加相位编码步数增加)为节省时间，可降低为节省时间，可降低Z方向的空间分辨方向的空间分辨力力. The effect of gradient fields on MR signal? Compare to SE signal with no gradient fields . 上次布置作业上次布置作业l图像重建作业 headraw.matl傅里叶逆变换重建图像，l将数据第1到64行数据置0，然后重建l将数据1到64行与193到256行置0，…l将数据奇数行置0， …l将数据偶数行置0， …l将数据放到512*512大小的矩阵中心，其他位置处为0， …. Image Characteristics. Which is better？What is the difference？Contrast. Which is better？What is the difference？Resolution. Which is better？What is the difference？ Noise & SNR. Which is better？What is the difference？Artifact. Image QualityContrastResolutionNoise, Signal-to-Noise RatioArtifact. Image QualityContrastResolutionNoise, Signal-to-Noise RatioArtifactT1-weighted, T2-weighted, Proton Density. At a main field of 1.5 TTissue TypeApproximate T1 value in msApproximate T2 value in msAdipose tissues240-25060-80Whole blood (deoxygenated)135050Whole blood (oxygenated)1350200Cerebrospinal fluid (similar to pure water)4200 - 45002100-2300Gray matter of cerebrum920100White matter of cerebrum78090Liver49040Kidneys65060-75Muscles860-90050In most situations (but not in principle) T1 is greater than T2. . PDW T1W T2WContrast. MR 基本脉冲序列基本脉冲序列l饱和恢复序列饱和恢复序列 ~ Saturation recovery SRl部分饱和恢复序列部分饱和恢复序列 ~ Partial SRl反转恢复序列反转恢复序列 ~ Inversion recovery IRlStandard SElGRE. 饱和恢复序列SaturationRecovery1. What is Saturation?In MR, saturation is a nonequilibrium state with no net magnetizationThe same amount of nuclear spins is aligned against and with the magnetic field2. How？？Saturation condition？？TR >> T1 OR T2 ？？3T2. 饱和恢复序列饱和恢复序列Saturation Recovery. The amplitude of the FID signal is ？？:. 饱和恢复序列SaturationRecoveryl每个每个90RF后，产生一个后，产生一个FID信号信号l如果如果M从饱和和态完全恢复，每个从饱和和态完全恢复，每个FID都可都可达到最大值达到最大值l90 - (长间隔长间隔) – 90l . II. 部分饱和恢复序列部分饱和恢复序列 Partial SRl90-(短间隔短间隔)- 90l第第2个个90°RF脉冲后，脉冲后，M没有完全恢复，后面的没有完全恢复，后面的FID信号比第信号比第1个个FID信号小。
信号小The effect of LongTR on Af？？. III. 反转恢复序列反转恢复序列 IR ~ Inversion Recoveryl180°- 90 °l在在IR序列中，首先使用序列中，首先使用180°RF 脉冲，然脉冲，然后等待一段时间再使用一个后等待一段时间再使用一个 90°RF 脉冲 III. 反转恢复序列反转恢复序列 IR ~ Inversion RecoveryTI( Inversion Time). l对于典型的对于典型的SR序列序列T1 恢复曲线，纵向恢复曲线，纵向弛豫恢复曲线：弛豫恢复曲线： l1-exp(-t/T1)l在在IR序列中，序列中，M由由–M0 恢复到恢复到M0，，恢复曲线：恢复曲线：l1-2exp(-t/T1). Signal Nulling effectl信号过信号过0的点叫做的点叫做null pointl过过0点的时间点的时间TI(null)l TI(null)l = (ln2)T1l = 0.693T1. 脂肪抑制脂肪抑制~ Fat suppression STIR （（Short TI Inversion Recovery））l选择选择TI使脂肪信号过使脂肪信号过0点点l其他组织：其他组织：M可进入可进入xy平面产生随平面产生随T2衰减的衰减的信号；信号；l脂肪：没有脂肪：没有M可以可以l 进入进入xy平面；平面；l1.5T：： TI ~ 140msl0.5T：： TI ~ 100ms. Radiotherapy induced fatty marrowA 67 y o female , a known case of Ca cervix with radiotherapy. Now complains of severe backache, radiating to lower limbs. Referred to MRI lumbar spine to rule out metastasis and cause of pain. Diffuse homogeneously high marrow signals involving L5 and all sacral vertebral bodies on T1 and T2 w images with complete signal suppression on STIR implies to fatty marrow, it s a radiotherapy induced change, area of involvement corresponds to field of radiotherapy, should not be mistaken for neoplastic marrow infiltration. . （液体〕流体抑制（液体〕流体抑制 FLAIR（（ Fluid-Attenuated Inversion Recovery ））l选择选择TI使使water信号过信号过0点点l其他组织：其他组织：M可进入可进入xy平面产生随平面产生随T2衰减的衰减的信号；信号；lwater ：没有：没有M可以可以l 进入进入oxy平面；平面；lTI：：2000ms. . IV. Basic Spin-Echo Imaging. SE ~ Spin echo sliceechosignalTERFGzreadoutphaseGyGx. l1：：90选择性选择性RF脉冲与脉冲与 Gz —层面选择层面选择l2：：-Gz ：产生反方向相移，部分补偿：产生反方向相移，部分补偿+Gz产生的正向相移产生的正向相移—提高信号提高信号l3：：Gy — 相位编码相位编码l4:180 RF 脉冲脉冲— 形成回波形成回波l5:-Gx 产生负向相移，部分补偿后面产生负向相移，部分补偿后面 + Gx 产生的正向相移产生的正向相移—采集对称信号采集对称信号l6:+Gx —频率编码〔采样、读数）频率编码〔采样、读数）. echosignalTERFTR. Saturation-Recovery Spin-Echo. Inversion-Recovery Spin-Echo. . TRTEPDWLongShortT1WShortShortT2WLongLong？？ShortLong一般地，短TR：100—1000ms长TR：1600—3000ms短TE：10—30ms长TE：70—100ms. cis.rit.edu/htbooks/mri/inside.htmTR ValueTE ValueT1W<= T1<< T2T2W>> T1>=T2PDW>>T1<< T2. 图像 A: TR = 1500 ms B: TR = 2500 ms C: TR= 4000 TE = 30PD像像:改变改变TR. Contrast. cis.rit.edu/htbooks/mri/inside.htm. 图像 A: TR = 350 图像 B: TR = 500 图像 C: TR= 1000T1像像:改变改变TR. 图像 A: TE = 9 图像B: TE = 14 图像 C: TE = 40T1像像:改变改变TE. Contrast. cis.rit.edu/htbooks/mri/inside.htm. T2像像:改变改变TEA: TE = 25B: TE = 50C: TE = 75D: TE = 100. 图像A: TR= 1500 B: TR= 2500 C: TR= 4000 TE = 90T2像像:改变改变TR. •Normal tissue usually have a small variation in spin density but have quite different T1values •T1W image usually have a good anatomical definition•Many disease states are characterized by a change of the tissue T2 value,•T2W imaging is a sensitive method for disease detection. 右侧小脑半球脑膜瘤的病人，T1WI因为病变呈等信号，所以相对来说比较不容易发现病变；T2WI病变的信号异常明显，可见病变中间信号较高〔较亮），病变周边信号较低〔较暗），比较利于病变的显示。
Basic Gradient EchoThe T1 contrast is dependent mainly on the FA instead of the TRT2* contrast depends on TET1 contrastTRFA ?cis.rit.edu/htbooks/mri/inside.htm. 一般的：一般的：SE的的SNR比比GRE的高的高GRE的伪影比的伪影比SE的多的多GRE的速度比的速度比SE的快的快SE得到得到T2 对比，对比，GRE得到得到T2*对比对比 SE的图像质量比的图像质量比GRE的好的好,但速度慢但速度慢 GRE vs SEEcho formation; lower FA. Image QualityContrastResolutionNoise, Signal-to-Noise RatioArtifact. Resolution. Resolution. 空间分辨力空间分辨力l视野视野Field of View，， FOVl空间分辨力与空间分辨力与 (FOV)和扫描矩阵的关系：和扫描矩阵的关系：l x = FOVx / Nxl y = FOVy / NyFOVxFOVyFOVx，，FOVy =240mm，，Nx = Ny =256，求，求 x，， y. Resolution. How to Set FOV and Resolution in MRI. kykxIDFTk space sampling and imageyx. •FOV中央，中央，B0•FOV右边，大于右边，大于 B0•FOV左边，小于左边，小于B0 x= Bx= Gxx fmax= GxFOVx/2or BW= GxFOVxBandWidth BW=2 fmax . FOV=BW/ G=1/ G Ts 1/FOV= G Ts，， Ts :dwell timeX方向：方向： 1/FOVx= Gx txBW= GxFOVx. 定义定义 Gx tx为为 kx kx = Gx tx=1/FOVx kx x=1/Nx1/FOVx= Gx txy方向？方向？. 降低降低 FOV 的方法？的方法？:Stronger gradientLower BWFOV=BW/ G提高提高Resolution的方法？的方法？. 上次布置作业上次布置作业l图像重建作业 headraw.matl傅里叶逆变换重建图像，l将数据第1到64行数据置0，然后重建l将数据1到64行与193到256行置0，…l将数据奇数行置0， …l将数据偶数行置0， …l将数据放到512*512大小的矩阵中心，其他位置处为0， …. kykxFT-1K空间数据：复数空间数据：复数. k 空间特点空间特点k空间实际上是空间实际上是频率空间频率空间K空间位置与信空间位置与信号的空间位置没号的空间位置没有直接的对应关有直接的对应关系；系；k 空间内每个数空间内每个数据点都对图像有据点都对图像有贡献贡献k空间中心空间中心. . . 上次布置作业上次布置作业l图像重建作业 headraw.matl傅里叶逆变换重建图像，l将数据第1到64行数据置0，然后重建l将数据1到64行与193到256行置0，…l将数据奇数行置0， …l将数据偶数行置0， …l将数据放到512*512大小的矩阵中心，其他位置处为0， …. Noise， SNR. 信噪比Signal-to-Noise Ratio SNRlSNR信号水平与噪声标准差的比值l影响因素l分辨率高，每个像素或体素越小，SNR下降l接收带宽越高，noise水平越高，SNR下降l提高方法l多次采集平均可提高SNR ，但成像变慢l（Average or NEX：Number of EXcitation）l图像后处理去噪 Image postprocessing. Two Bandwidth？lTransmitlReceiver. Transmit BandwidthlThe transmit bandwidth refers to the RF excitation pulse required for slice selection in a pulse sequence. The slice thickness is proportional to the bandwidth of the RF pulse (and inversely proportional to the applied gradient strength). Lowering the pulse bandwidth can reduce the slice thickness.. lThe receiver (or acquisition) bandwidth (rBW) is the range of frequencies accepted by the receiver to sample the MR signal. lThe receiver bandwidth is changeable and has a direct relationship to the signal to noise ratio (SNR)l SNR = 1/squareroot (rBW)lrBW=GxFOVx lrBW=1/Ts (dwell time)lSometimes: rBW is defined by BW = Sampling Rate/Number of Samples (the bandwidth per pixel)Receiver Bandwidth. lA smaller bandwidth improves SNR, but can cause spatial distortions, also increases the chemical shift. A larger bandwidth reduces SNR (more noise from the outskirts of the spectrum), but allows faster imaging.. Noise in MR imagingNEX=1NEX=2NEX=4. MR图像去噪研究背景：临床常用图像信噪比基本能接受然而，还有很多情况成像的信噪比仍然较低—高b值DWI成像—高分辨率成像—动态成像 b=1000 b=2000 b=3000. 基于高阶奇异值分解的磁共振图像噪声去除算法TruthNoisy image (15%)PRI-NLM3DBM4DHOSVD-RMR图像去噪Xinyuan Zhang, Yanqiu Feng*, et al., Medical Image Analysis, 2019.. 基于高阶奇异值分解的磁共振图像噪声去除算法Xinyuan Zhang, Yanqiu Feng*, et al., Medical Image Analysis, 2019.MR图像去噪. 基于高阶奇异值分解的磁共振图像噪声去除算法MR图像去噪Xinyuan Zhang, Yanqiu Feng*, et al., Medical Image Analysis, 2019.. DWI图像噪声去除Xinyuan Zhang, Yanqiu Feng*, et al., Submitted to Neuroimage, 2019.b=3000 b=2000 b=1000LR+EdgeNoisyGL-HOSVDNLM. FAMapXinyuan Zhang, Yanqiu Feng*, et al., Submitted to Neuroimage, 2019.LR+EdgeNoisyGL-HOSVDNLMb=3000 b=2000 b=1000DWI图像噪声去除. NoisyGL-HOSVDNLMNEX=10DW imageFA mappingEnlarged regionDWI图像噪声去除Xinyuan Zhang, Yanqiu Feng*, et al., Submitted to Neuroimage, 2019.. Image QualityContrastResolutionNoise, Signal-to-Noise RatioArtifact （（by Teacher Lili Kang））. 。