潘輝王亮 王強(qiáng)龍 陳利民賈峰 劉震宇?
1)(中國(guó)科學(xué)院長(zhǎng)春光學(xué)精密機(jī)械與物理研究所,長(zhǎng)春 130033)2)(中國(guó)科學(xué)院大學(xué),北京 100039)3)(德國(guó)弗萊堡大學(xué)醫(yī)學(xué)院醫(yī)學(xué)中心放射學(xué)系醫(yī)學(xué)物理科,弗萊堡 79110)
基于Pareto優(yōu)化理論的多目標(biāo)超橢梯度線圈設(shè)計(jì)?
潘輝1)2)王亮1)王強(qiáng)龍1)陳利民1)2)賈峰3)劉震宇1)?
1)(中國(guó)科學(xué)院長(zhǎng)春光學(xué)精密機(jī)械與物理研究所,長(zhǎng)春 130033)2)(中國(guó)科學(xué)院大學(xué),北京 100039)3)(德國(guó)弗萊堡大學(xué)醫(yī)學(xué)院醫(yī)學(xué)中心放射學(xué)系醫(yī)學(xué)物理科,弗萊堡 79110)
(2016年12月21日收到;2017年2月6日收到修改稿)
磁共振系統(tǒng)梯度線圈設(shè)計(jì)是一個(gè)多目標(biāo)優(yōu)化問(wèn)題,在設(shè)計(jì)時(shí)需要綜合考慮能耗、磁場(chǎng)能、線性度等設(shè)計(jì)要求.這些設(shè)計(jì)要求通常難以同時(shí)獲得極小解,因此在設(shè)計(jì)梯度線圈時(shí)需要權(quán)衡線圈的各方面的設(shè)計(jì)需求.本文基于柱面可展性和流函數(shù)設(shè)計(jì)方法,結(jié)合Pareto優(yōu)化方法實(shí)現(xiàn)了在超橢圓柱設(shè)計(jì)表面上梯度線圈的多目標(biāo)設(shè)計(jì).分別分析了磁場(chǎng)能、能耗目標(biāo)對(duì)梯度線圈線性度、線圈構(gòu)型的影響;并在Pareto解空間中分析各目標(biāo)的相互變化關(guān)系,通過(guò)數(shù)值算例驗(yàn)證了該方法在超橢梯度線圈設(shè)計(jì)時(shí)的有效性與靈活性.優(yōu)化結(jié)果顯示,在滿足線性度誤差小于5%,能耗與磁場(chǎng)能分別小于用戶設(shè)定值的設(shè)計(jì)約束下,梯度線圈的多目標(biāo)設(shè)計(jì)存在多個(gè)局部?jī)?yōu)化解.該方法可以直觀地比較相同目標(biāo)函數(shù)值的情況下各單目標(biāo)的具體表現(xiàn),有利于實(shí)現(xiàn)不同的設(shè)計(jì)要求下梯度線圈的最終定型設(shè)計(jì).
梯度線圈,流函數(shù),超橢圓柱面,Pareto優(yōu)化
磁共振成像作為一種非介入式影像技術(shù),能夠反映物體內(nèi)部的層次結(jié)構(gòu).在過(guò)去的幾十年,磁共振成像已經(jīng)成為醫(yī)療診斷、生物研究、材料研究等領(lǐng)域的重要途徑之一.磁共振成像是利用預(yù)先定義的可控磁場(chǎng)(如線性梯度磁場(chǎng))來(lái)編碼信號(hào)的空間位置信息[1].因此磁場(chǎng)的線性度、均勻度、強(qiáng)度將直接影響設(shè)備成像的效果.梯度磁場(chǎng)由線圈產(chǎn)生,通常布置在圓柱表面上,但為了更好地滿足醫(yī)療需求,本文采用在超橢圓柱表面上設(shè)計(jì)梯度線圈的方法,以期減少被檢測(cè)者的幽閉恐懼癥[2].
現(xiàn)有梯度線圈的主要研究方法有目標(biāo)場(chǎng)法和流函數(shù)法.目標(biāo)場(chǎng)法在1986年由Tuner首先提出,結(jié)合Fourier變換求解電流密度[3,4].在此基礎(chǔ)上,該方法被應(yīng)用到雙平面、圓柱面等設(shè)計(jì)表面;文獻(xiàn)[5—9]進(jìn)一步發(fā)展了圓柱面等簡(jiǎn)單規(guī)則設(shè)計(jì)曲面的目標(biāo)場(chǎng)法.流函數(shù)法[10?13]是基于有限元離散求解的思想,更適應(yīng)于復(fù)雜曲面的梯度線圈設(shè)計(jì).現(xiàn)有的算法實(shí)現(xiàn)方式主要有直接法、鏡像法和有限元法[14];王秋良課題組[15,16]在此基礎(chǔ)上進(jìn)一步實(shí)現(xiàn)了算法的改進(jìn)與快速計(jì)算.近二十幾年的發(fā)展中,梯度線圈的設(shè)計(jì)研究在磁場(chǎng)線性度、均勻性方面有了很大的進(jìn)展[17,18],但成像響應(yīng)時(shí)間、能耗、振動(dòng)與噪音等性能指標(biāo)仍需要進(jìn)一步優(yōu)化,從而提高儀器的效率.響應(yīng)時(shí)間與電感直接相關(guān),降低線圈總電感值不僅可以減少能耗,還可以減小磁場(chǎng)的渦流效應(yīng).本文利用流函數(shù)法以梯度線圈的電感值、能耗等為優(yōu)化目標(biāo)進(jìn)行設(shè)計(jì)研究,從而減小線圈響應(yīng)時(shí)間,降低能耗.
為了提供更好的性能和患者使用時(shí)的舒適度,理想的線圈同時(shí)具有最小的磁場(chǎng)能、最高的電磁轉(zhuǎn)化率、最小的電阻、較好的磁場(chǎng)線性度、較短的響應(yīng)時(shí)間以及較低的由洛倫茲力引起的噪音等性能.然而,該問(wèn)題并不存在一個(gè)同時(shí)使所有目標(biāo)性能最優(yōu)的解,而是可能存在一系列權(quán)衡各性能的局部最優(yōu)解.因此需要同時(shí)優(yōu)化多個(gè)目標(biāo)來(lái)獲取更好的綜合性能.另一方面,梯度線圈設(shè)計(jì)問(wèn)題屬于多目標(biāo)問(wèn)題[17?25],以單個(gè)磁場(chǎng)線性度為目標(biāo)來(lái)求解具有不存在惟一解、線型振蕩等問(wèn)題.之前的研究中,為避免這類問(wèn)題采用Tikhonov正則化方法構(gòu)建目標(biāo)函數(shù)[2],但該方法并無(wú)對(duì)應(yīng)的實(shí)際物理意義.本文結(jié)合Pareto優(yōu)化理論[23]對(duì)磁共振系統(tǒng)梯度線圈的線性度、能耗、磁場(chǎng)能等多目標(biāo)進(jìn)行優(yōu)化分析.
目前,磁共振梯度線圈的設(shè)計(jì)方法主要可分為目標(biāo)場(chǎng)法和流函數(shù)法.相對(duì)于目標(biāo)場(chǎng)法,流函數(shù)法具有更加靈活、適應(yīng)復(fù)雜設(shè)計(jì)曲面的優(yōu)勢(shì).本文采用流函數(shù)法進(jìn)行分析計(jì)算.梯度線圈設(shè)計(jì)是一個(gè)多目標(biāo)優(yōu)化問(wèn)題,需綜合考慮磁場(chǎng)的線性度、磁場(chǎng)能、線圈的能耗等因素.
2.1流函數(shù)法
流函數(shù)法相對(duì)于目標(biāo)場(chǎng)法將電流密度的矢量求解轉(zhuǎn)化為流函數(shù)ψ的標(biāo)量求解,簡(jiǎn)化了求解過(guò)程,更適應(yīng)于超橢等復(fù)雜曲面的線圈設(shè)計(jì).故本文采用流函數(shù)法在超橢圓柱設(shè)計(jì)表面(圖1)上進(jìn)行線圈的優(yōu)化設(shè)計(jì),電流密度J可表示為
其中,ne= (nex,ney,0)是超橢設(shè)計(jì)表面[2](x/a)2/m+(y/b)2/n=1的單位外法向量;nex,ney分別是單位外法向量ne的x,y方向分量;a,b分別是超橢圓的長(zhǎng)短半軸的半軸長(zhǎng);m,n∈(0,1].
圖1 超橢設(shè)計(jì)表面與ROI及設(shè)計(jì)表面的展開(kāi)面Fig.1.Current-carrying surface Γcoil,ROI,and the developed design surface.
根據(jù)Biot-Savart定律,目標(biāo)區(qū)域(region of interest,ROI)內(nèi)第i點(diǎn)的磁場(chǎng)強(qiáng)度z方向分量可表示為
其中,
Zu,Zl分別是超橢圓柱面Z軸向的上下邊界;(ri,θi,zi)是ROI的場(chǎng)點(diǎn)坐標(biāo),(θ,z)是超橢設(shè)計(jì)表面展開(kāi)坐標(biāo).
利用有限單元法分片逼近的思想[2],超橢圓柱面展開(kāi)平面上的流函數(shù)ψ可表示為
其中φj為各節(jié)點(diǎn)值,Nj為插值函數(shù),v是設(shè)計(jì)表面Γcoil上各節(jié)點(diǎn)的自由度總和.
2.2梯度線圈設(shè)計(jì)目標(biāo)和約束
本文中根據(jù)梯度線圈的性能的重要性以及病人的舒適度,在超橢圓柱設(shè)計(jì)表面上對(duì)梯度線圈的線性度、能耗和磁場(chǎng)能進(jìn)行了多目標(biāo)優(yōu)化.
2.2.1 磁場(chǎng)強(qiáng)度與線性度
磁共振系統(tǒng)成像的原理在于利用線性的梯度磁場(chǎng)Gx,Gy,Gz來(lái)區(qū)分不同空間位置的磁共振信號(hào),實(shí)現(xiàn)空間定位[1].梯度線圈產(chǎn)生的磁場(chǎng)分布逼近目標(biāo)梯度磁場(chǎng)分布是實(shí)現(xiàn)其定位功能的基本要求.因此,在已有的梯度線圈優(yōu)化模型中,都將ROI內(nèi)真實(shí)磁場(chǎng)強(qiáng)度z分量的真實(shí)值Bz與理想值的偏離值和作為優(yōu)化的主要設(shè)計(jì)目標(biāo),可表達(dá)為
磁場(chǎng)線性度在工程中是衡量線圈成像質(zhì)量的一個(gè)重要參數(shù).梯度磁場(chǎng)的線性度將直接影響到成像的質(zhì)量,線性度越高,空間定位越精確,獲得的圖像質(zhì)量就越好.磁場(chǎng)線性度可用最大線性誤差gΔBz來(lái)衡量:
2.2.2 梯度線圈能耗
梯度磁場(chǎng)需由梯度線圈通過(guò)電流驅(qū)動(dòng)產(chǎn)生,在此過(guò)程中會(huì)由于線圈的電阻產(chǎn)生大量的熱量,造成設(shè)備過(guò)熱、能源消耗過(guò)多等.因此在設(shè)計(jì)優(yōu)化中,將考慮由于線圈電阻產(chǎn)生的能耗,從而提高線圈的電磁轉(zhuǎn)換率.在實(shí)際計(jì)算中,線圈的能耗可表示為
其中,σ為電導(dǎo)率,t是導(dǎo)電層厚度.
根據(jù)能耗fP與電阻R、電流I的關(guān)系,可以求出表征線圈性能的重要參數(shù)之一線圈的電阻R.
其中I線圈電流,計(jì)算中一般取I=(ψmax?ψmin)/N,ψmax和ψmin分別是流函數(shù)的最大值和最小值[11],N是線圈圈數(shù).
2.2.3 梯度線圈磁場(chǎng)能
在磁共振設(shè)備工作時(shí),需頻繁切換線圈的電流來(lái)達(dá)到成像定位的作用.因此線圈的電感將直接影響到設(shè)備的響應(yīng)速度與成像時(shí)間.而磁場(chǎng)能的大小反映了線圈的電感值,優(yōu)化磁場(chǎng)能將等效于優(yōu)化線圈電感[4,19].
磁場(chǎng)能與電感的關(guān)系表達(dá)式:
在該問(wèn)題中磁場(chǎng)能可由(9)式求得
其中,A是磁位移矢量.由于在非設(shè)計(jì)區(qū)域J=0,所以磁場(chǎng)能可寫(xiě)成
2.3多目標(biāo)優(yōu)化
2.3.1 Pareto多目標(biāo)優(yōu)化
實(shí)際工程問(wèn)題的優(yōu)化大多基于多約束條件下的多目標(biāo)優(yōu)化問(wèn)題.對(duì)于某些等式或不等式約束條件,可轉(zhuǎn)化為輔助優(yōu)化目標(biāo).要得到滿足各要求的優(yōu)化解,需對(duì)各個(gè)目標(biāo)、約束進(jìn)行權(quán)衡并建立適于求解的多目標(biāo)函數(shù)形式[26?28].多目標(biāo)的一般函數(shù)形式為
(11)式中Fk(x)為第k個(gè)子目標(biāo)函數(shù),gj(x)為第j個(gè)不等式約束,hl(x)為第l個(gè)等式約束.Pareto優(yōu)化[23]是指有且僅有x?∈X 使得對(duì)于x∈X有F(x)≤ F(x?),并且至少存在一個(gè)目標(biāo)函數(shù)有Fi(x)≤ Fi(x?),則點(diǎn)x?為Pareto優(yōu)化. 當(dāng)有且僅有x?∈X對(duì)于x∈X不再存在x使得有F(x) ≤ F(x?),則點(diǎn)x?為弱Pareto優(yōu)化. 若有Fo∈ Zk,如果對(duì)于每一個(gè)i=1,2,...,k,都存在則x為理想點(diǎn).
多目標(biāo)優(yōu)化有一些用來(lái)判斷評(píng)定優(yōu)化方法最終效果的概念,最常用的兩種概念是多目標(biāo)優(yōu)化必要條件和充分條件.如果一個(gè)多目標(biāo)優(yōu)化形式提供一個(gè)必要條件,對(duì)于一個(gè)Pareto優(yōu)化點(diǎn),則必為此多目標(biāo)優(yōu)化形式的解.對(duì)于提供必要條件的多目標(biāo)優(yōu)化形式,可通過(guò)調(diào)整參數(shù)來(lái)獲得所有的Pareto優(yōu)化點(diǎn),如果一個(gè)多目標(biāo)優(yōu)化形式提供一個(gè)充分條件,則其解為Pareto優(yōu)化,但是一個(gè)確定的Pareto優(yōu)化點(diǎn)可能無(wú)法獲取.
2.3.2 梯度線圈的多目標(biāo)優(yōu)化
梯度線圈的優(yōu)化需要綜合考慮上述線性度、能耗、磁場(chǎng)能等目標(biāo)及約束[13],可表達(dá)為
多目標(biāo)處理方法主要有:全局加權(quán)準(zhǔn)則法、加權(quán)和法、字典式法、加權(quán)最小最大法、指數(shù)加權(quán)法和加權(quán)積法等.為了更好地分析各個(gè)目標(biāo)參數(shù),本文采用加權(quán)和與歸一化相結(jié)合的多目標(biāo)處理方法[24,25],其表達(dá)形式為
ωB,ωP,ωE分別是目標(biāo)函數(shù)fB,fP,fE的對(duì)應(yīng)權(quán)重系數(shù).
2.3.3 敏度分析
在優(yōu)化過(guò)程中,需要計(jì)算函數(shù)的敏度:
根據(jù)(2)和(4)式可求出線性度函數(shù)fB的敏度:
其中p是ROI離散點(diǎn)的總數(shù).根據(jù)(6)式,能耗目標(biāo)函數(shù)的流函數(shù)表達(dá)形式可寫(xiě)成
對(duì)其求敏度
根據(jù)(10)式可推出磁場(chǎng)能目標(biāo)可表示為
對(duì)于加權(quán)目標(biāo)f(φ),其敏度可由(14)式計(jì)算.
本文以y方向梯度線圈設(shè)計(jì)為例在超橢圓柱設(shè)計(jì)表面進(jìn)行多目標(biāo)梯度線圈設(shè)計(jì).設(shè)定ROI磁場(chǎng)強(qiáng)度梯度的目標(biāo)值
3.1結(jié)合磁場(chǎng)能的優(yōu)化
本小節(jié)只討論磁場(chǎng)能與磁場(chǎng)線性度,故能耗的權(quán)重系數(shù)ωP取為0,目標(biāo)函數(shù)可表達(dá)為
其中ωB+ωE=1.
為方便比較,計(jì)算中以優(yōu)化后的線圈Coil0為參照進(jìn)行對(duì)比,其中Coil0是磁場(chǎng)能權(quán)重系數(shù)ωE=4.8×10?4時(shí)的計(jì)算結(jié)果,參數(shù)如表1所列.
由圖2(b)可知在Coil0處有
表1 梯度線圈Coil0參數(shù)Table 1.Parameters of the gradient coil,Coil0.
如圖2所示,增加磁場(chǎng)能目標(biāo)的權(quán)重值能夠有效降低磁場(chǎng)能,但隨著磁場(chǎng)能權(quán)重系數(shù)的增加將導(dǎo)致線性度目標(biāo)fB的增加,降低了線性度.由(21)式可知在Coil0處磁場(chǎng)能降低1%將導(dǎo)致磁場(chǎng)線性度損失10.11%;然而磁場(chǎng)能與電感呈線性關(guān)系,增加磁場(chǎng)能將導(dǎo)致電感增加,進(jìn)一步會(huì)增加磁共振系統(tǒng)的響應(yīng)時(shí)間.因此,在設(shè)計(jì)中需綜合考慮各因素,針對(duì)設(shè)計(jì)需求進(jìn)行優(yōu)化設(shè)計(jì).
磁場(chǎng)能目標(biāo)除了能夠優(yōu)化磁場(chǎng)能的大小外,還可以用來(lái)優(yōu)化線圈的設(shè)計(jì)尺寸.增加ROI的體積將導(dǎo)致磁場(chǎng)能增加.在ROI尺寸不變時(shí),合理地降低線圈分布表面尺寸,將提高磁共振系統(tǒng)的空間利用率,提高磁共振系統(tǒng)的工作效率.在本文中,利用優(yōu)化磁場(chǎng)能來(lái)得到合理的線圈分布表面尺寸.
如圖3所示,隨著線圈分布表面高度h的增加,線性度與磁場(chǎng)能均減小,在h增加到180 mm時(shí),磁場(chǎng)能與場(chǎng)強(qiáng)線性度趨于平緩;當(dāng)h增大至400 mm時(shí)增加高度會(huì)導(dǎo)致磁場(chǎng)能變大而線性度變好.對(duì)比圖4中在不同高度、相同權(quán)重系數(shù)下的線圈線型,可以發(fā)現(xiàn)降低h會(huì)存在線型振蕩,線圈局部密集的現(xiàn)象;而h增大至270 mm后繼續(xù)增加h值獲得的梯度線圈線型的高度相同.因此,在后續(xù)的計(jì)算中,本文都采用h=270 mm設(shè)計(jì)分析.通過(guò)以上分析可以發(fā)現(xiàn)結(jié)合磁場(chǎng)能目標(biāo)與線圈的線型可以更好地選取滿足設(shè)計(jì)要求的線圈尺寸參數(shù).
3.2結(jié)合能耗的優(yōu)化
本小節(jié)將討論線圈能耗與磁場(chǎng)線性度對(duì)線型的影響,磁場(chǎng)能的權(quán)重系數(shù)ωE取為0,目標(biāo)函數(shù)可表達(dá)為
其中ωB+ωP=1.
圖2 (a)歸一化fB-fE的Pareto曲線及其(b)對(duì)應(yīng)的偏微分曲線Fig.2.(a)Pareto curve in the normalized fB-fEand(b)corresponding partial di ff erential curve.
圖3 設(shè)計(jì)表面高度h與(a)梯度線圈線性度、(b)磁場(chǎng)能線圈的關(guān)系(計(jì)算時(shí)取ωE=4.8×10?4)Fig.3.The relationship of the height of current-carrying surface to(a)normalized objection fB/fB0and(b)normalized objection fE/fE0(weight parameter:ωE=4.8×10?4).
圖4 (a)—(e)h=150,200,270,300和350 mm時(shí)的四分之一線型Fig.4.Quadrant of wire paths of gradient coils corresponding to(a)h=150 mm,(b)h=200 mm,(c)h=270 mm,(d)h=300 mm,(e)h=350 mm.
線圈的能耗主要受線圈的長(zhǎng)度與電流的大小影響,優(yōu)化能耗函數(shù),其實(shí)質(zhì)是對(duì)線圈線長(zhǎng)的約束,使線型規(guī)整,減少線型振蕩的產(chǎn)生.為方便比較能耗目標(biāo)對(duì)梯度線圈設(shè)計(jì)的影響,該部分的設(shè)計(jì)表面都設(shè)定為a=45 mm,b=36 mm,h=270 mm.
如圖5所示,通過(guò)增加能耗目標(biāo)權(quán)重系數(shù)可以有效減小梯度線圈的能耗.根據(jù)實(shí)際對(duì)線性度與能耗需求可以選取合適的權(quán)重系數(shù),使梯度磁場(chǎng)的線性度控制在合理的范圍內(nèi).對(duì)比圖中A,B位置的參數(shù)與線型如表2和圖6所示,發(fā)現(xiàn)增大能耗的權(quán)重系數(shù),可以減小線圈電流,同時(shí)使線圈更光滑.
圖5 磁場(chǎng)線性度與線圈能耗目標(biāo)值和權(quán)重系數(shù)的關(guān)系Fig.5.The relationship of the weight parameter for power dissipated ωPto objective value for power dissipated fPand objective value for linear gradient deviation fB.
3.3結(jié)合能耗、磁場(chǎng)能的多目標(biāo)優(yōu)化
在實(shí)際應(yīng)用中,梯度線圈的設(shè)計(jì)是需要綜合考慮多個(gè)性能參數(shù)的多目標(biāo)優(yōu)化問(wèn)題.本節(jié)將討論綜合磁場(chǎng)能、能耗、線性度目標(biāo)設(shè)計(jì)梯度線圈,目標(biāo)函數(shù)表達(dá)形式為(13)式.
表2 圖5中A,B位置梯度線圈的性能參數(shù)Table 2.Parameters of gradient coils of coil solutions noted as A,B in Fig.5.
圖6 (網(wǎng)刊彩色)(a),(b)分別對(duì)于圖5中A,B位置的線型;(c),(d)是(a),(b)對(duì)應(yīng)的局部線型Fig.6.(color online)(a)Wire paths and stream function of coil solutions noted as A in Fig.5,and(c)part of wire paths corresponding to(a);(b)wire paths and stream function of coil solutions noted as B in Fig.5,and(d)part of wire paths corresponding to(b).
圖7 (網(wǎng)刊彩色)Pareto解空間(滿足gΔBz≤ 5%,fE≤ 1.25×10?4,fP≤ 0.105),其中(a),(b),(c)分別是對(duì)應(yīng)fP,fE,fB的值Fig.7.(color online)Pareto front solution of multiple objectives optimization problem de fi ned by the objectives fB,fE,fP,and constrained to gΔBz≤ 5%,fE≤ 1.25×10?4,fP≤ 0.105:(a)corresponding to fP,(b)corresponding to fE,(c)corresponding to fB.
表3 圖7(c)中C,D,E位置Pareto優(yōu)化解對(duì)應(yīng)梯度線圈的參數(shù)Table 3.Parameters of gradient coils of Pareto front solutions noted as C,D,E in Fig.7(c).
圖8 (網(wǎng)刊彩色)圖7中C,D,E位置線型比較(黑色虛線是D點(diǎn)的線型,藍(lán)色是E點(diǎn)的線型,紅色是C點(diǎn)的線型)Fig.8.(color online)Quadrant of wire paths of gradient coils of Pareto front solutions noted as C,D,E inFig.7(c):(a)coil solutions noted as C(in red solid line),coil solutions noted as D(in black dotted line),(b)coil solutions noted as D(in black dotted line),coil solutions noted as E(in blue solid line).
圖7中三角形區(qū)域是滿足gΔBz≤5%,fP≤1.25×10?4,fE≤ 0.105的Pareto解空間.通過(guò)比較圖7(a)—(c)的變化趨勢(shì),根據(jù)實(shí)際需求可以直觀地得到可行解.表3為圖7(c)中C,D,E位置Pareto優(yōu)化解對(duì)應(yīng)梯度線圈的參數(shù).比較圖8中的線型可以發(fā)現(xiàn)增加能耗或磁場(chǎng)能的權(quán)重將會(huì)有效抑制線圈的局部振蕩.
梯度線圈的設(shè)計(jì)是一個(gè)多目標(biāo)優(yōu)化問(wèn)題,需要權(quán)衡各性能指標(biāo),如能耗、磁場(chǎng)能、線性度等.在形狀設(shè)計(jì)時(shí)也需要考慮磁場(chǎng)能等目標(biāo)來(lái)達(dá)到最佳線圈尺寸,從而提高磁共振系統(tǒng)的空間利用率和工作效率.在實(shí)際設(shè)計(jì)過(guò)程中,針對(duì)不同的應(yīng)用需求,需要性能不同的線圈.本文嘗試?yán)枚嗄繕?biāo)優(yōu)化方法實(shí)現(xiàn)超橢圓柱設(shè)計(jì)表面梯度線圈的設(shè)計(jì)問(wèn)題.通過(guò)計(jì)算表明該方法可以直觀地反映多個(gè)性能參數(shù)的關(guān)系,從而在設(shè)計(jì)過(guò)程中根據(jù)需求更好地權(quán)衡各個(gè)參數(shù).從多目標(biāo)優(yōu)化設(shè)計(jì)模型的角度考慮,文中采用的算法可以擴(kuò)展到多于四個(gè)合理目標(biāo)的線性組合.但是局部最優(yōu)解的存在與否強(qiáng)烈依賴于各個(gè)單目標(biāo)約束值的選取.如果約束選取的過(guò)緊,非常有可能發(fā)生可行解不存在的情況.因此,Pareto算法僅僅給出了一個(gè)尋找多目標(biāo)(尤其是多目標(biāo)之間有沖突的情況下)優(yōu)化解的一個(gè)尋優(yōu)策略,局部最優(yōu)解的最終確定還是強(qiáng)烈依賴于設(shè)計(jì)人員對(duì)于各單個(gè)目標(biāo)特性了解的基礎(chǔ)上對(duì)優(yōu)化參數(shù)的調(diào)整.文中通過(guò)優(yōu)化磁場(chǎng)能目標(biāo)實(shí)現(xiàn)了線圈設(shè)計(jì)表面尺寸的優(yōu)化,再通過(guò)對(duì)比能耗、磁場(chǎng)能、線性度目標(biāo)的變化趨勢(shì)來(lái)設(shè)計(jì)出滿足需求的梯度線圈.其他目標(biāo)如振動(dòng)、主動(dòng)屏蔽等將在后續(xù)研究中進(jìn)一步探討.
[1]Zu D L 2004Magnetic Resonance Imaging(Beijing:Higher Education Press)pp53–82(in Chinese)[俎棟林2004核磁共振成像學(xué)(北京:高等教育出版社)第53—82頁(yè)]
[2]Wang L,Cao Y H,Jia F,Liu Z Y 2014Acta Phys.Sin.63 238301(in Chinese)[王亮,曹英暉,賈峰,劉震宇2014物理學(xué)報(bào)63 238301]
[3]Turner R 1986J.Phys.D:Appl.Phys.19 147
[4]Turner R 1988J.Phys.E:Sci.Instrum.21 948
[5]Forbes L K,Crozier S 2002J.Phys.D:Appl.Phys.35 839
[6]Liu W T,Zu D L,Tang X 2010Chin.Phys.B19 018701
[7]Forbes L K,Brideson M A,Crozier S 2005IEEE Trans.Magn.41 2134
[8]Liu W T,Zu D L,Tang X,Guo H 2007J.Phys.D:Appl.Phys.40 4418
[9]Li X,Xie D X,Wang J M 2009IEEE Trans.Magn.45 1804
[10]Tomasi D 2001Magn.Reson.Med.45 505
[11]Peeren G N 2003J.Comput.Phys.191 305
[12]Lemdiasov R A,Ludwig R 2005Concepts Magn.Reson.B:Magn.Reson.Eng.26B 67
[13]Liu Z Y,Jia F,Hennig J,Korvink J G 2012IEEE Trans.Magn.48 1179
[14]Wang Q L 2013Practical Design of Magnetostatic Structure Using Numerical Simulation(Singapore:John Wiley&Sons)pp39–142
[15]Hu G L,Ni Z P,Wang Q L 2012IEEE Trans.Appl.Supercond.22 4900604
[16]Zhu X C,Wang Q L,Wang H S 2016Adv.Technol.Electral.Eng.Energ.35 43(in Chinese)[朱旭晨,王秋良,王厚生2016電工電能技術(shù)35 43]
[17]Li X,Xia L,Chen W F,Liu F,Crozier S,Xie D X 2011J.Magn.Reson.208 148
[18]Hu Y,Wang Q L,Li Y,Zhu X C,Niu C Q 2016Acta Phys.Sin.65 218301(in Chinese)[胡洋,王秋良,李毅,朱旭晨,牛超群2016物理學(xué)報(bào)65 218301]
[19]Turner R 1993Magn.Reson.Imag.11 903
[20]Abduljalil A M,Aletras A H,Robilaille P M L 1994Magn.Reson.Med.31 450
[21]Alsop D C,Connick T J 1996Magn.Reson.Med.35 875
[22]Pissanetzky S 1992Meas.Sci.Technol.3 667
[23]Bowtell R,Robyr P 1998J.Magn.Reson.131 286
[24]Wang L Q,Wang W M 2014Chin.Phys.B23 028703
[25]Sanchez C C,Pantoja M F,Poole M,Bretones A R 2012IEEE Trans.Magn.48 1967
[26]Marler R T,Arora J S 2004Struct.Multid.Optim.26 369
[27]Marler R T,Arora J S 2005Eng.Optim.37 551
[28]Xie D X,Sun X W,Bai B D,Yang S Y 2008IEEE Trans.Magn.44 1006
PACS:83.85.Fg,02.60.Cb,87.55.de,45.10.DbDOI:10.7498/aps.66.098301
Design of super-elliptical gradient coils based on multiple objective Pareto optimization method?
Pan Hui1)2)Wang Liang1)Wang Qiang-Long1)Chen Li-Min1)2)Jia Feng3)Liu Zhen-Yu1)?
1)(Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Sciences,Changchun 130033,China)2)(University of Chinese Academy of Sciences,Beijing 100039,China)3)(Department of Radiology,Medical Physics,Medical Center University of Freiburg,Faculty of Medicine,University of Freiburg,Freiburg 79110,Germany)
21 December 2016;revised manuscript
6 February 2017)
The design of gradient coils for a magnetic resonance imaging(MRI)system is a multiple objective optimization problem,which usually needs to deal with a couple of con fl icting design objectives,such as the stored magnetic energy,power consumption,and target linear gradient distribution.These design requirements usually con fl ict with each other,and there is no unique optimal solution which is capable of minimizing all objectives simultaneously.Therefore,the design of gradient coils needs to be optimized reasonably with the tradeo ffamong di ff erent design objectives.Based on the developable property of the super-elliptical cylindrical surface and the stream function design method,the multiple objective optimization problem is analyzed by using the Pareto optimization method in this paper.The e ff ect of proposed approach is illustrated by using the stream function method and three aforementioned coil design objectives are analyzed.The in fl uences of the stored magnetic energy and power consumption target on linearity of gradient coil and the con fi guration of coils are analyzed respectively.The suitable sizes of gradient coils are discussed by analyzing the change of the stored magnetic energy.A weighted sum method is employed to produce the optimal Pareto solutions,in which the multiple objective problem reduces into a single objective function through a weighted sum of all objectives.The quantitative relationship of each design requirement is analyzed in the Pareto solution space,where Pareto optimal solutions can be intuitively found by dealing efficiently with the tradeo ffamong di ff erent coil properties.Numerical examples of super-elliptical gradient coil solutions are provided to demonstrate the e ff ectiveness and versatility of the proposed method to design super-elliptical gradient coils with di ff erent coil requirements.The optimization results show that there are multiple available solutions in the convex Pareto solution space under the constraints that the linear gradient deviation is less than 5%and the magnetic stored energy and power dissipated are both no more than user-preset values.In the case that the values of summed objective functions are the same,the proposed method can intuitively see the performance of each individual target,thereby conducting to realizing the fi nal design of gradient coils under the di ff erent design requirements.With the proposed approach,coil designers can have a reasonable overview of gradient coil design about the achievable performances of some speci fi c properties and the competing or compatible relationships among coils properties.Therefore,a suitable design of the gradient coils for a given requirement of MRI application can be chosen reasonably.
gradient coils,stream function,super-elliptical cylindrical surface,Pareto optimization
10.7498/aps.66.098301
?國(guó)家自然科學(xué)基金(批準(zhǔn)號(hào):51675506,51275504)、吉林省科技發(fā)展計(jì)劃(批準(zhǔn)號(hào):20140519007JH)和歐洲研究理事會(huì)ERC啟動(dòng)基金RANGEmri 282345項(xiàng)目資助的課題.
?通信作者.E-mail:liuzy@ciomp.ac.cn
*Project supported by the National Natural Science Foundation of China(Grant Nos.51675506,51275504),the Science and Technology Development Plan of Jilin Province,China(Grant No.20140519007JH),and an European Research Council Starting Grant ‘RANGEmri’(Grant Agreement 282345).
?Corresponding author.E-mail:liuzy@ciomp.ac.cn