Transceive Phased-array Systems for Parallel MRI

AuthorYu Li
Thesis TitleTransceive Phased-array Systems for Parallel MRI
School, Centre or InstituteSchool of Information Technol and Elec Engineering
InstitutionThe University of Queensland
Publication date2010-12
Thesis typePhD Thesis
SupervisorStuart Crozier
Feng Liu
Total pages267
Total colour pages83
Total black and white pages184
Subjects09 Engineering
Abstract/SummaryAbstract This thesis focuses on the transceive phased array designs for magnetic resonance imaging (MRI). With the introduction of phased array technology in MRI, a wide variety of phased array coils have been constructed for human clinical applications or small animal studies. Compared to receive-only phased array coils, transceive phased-arrays combine the function of the transmit coils and receiver coils. They have the capability of generating a homogeneous magnetic field and detecting the MRI signal with a high signal-noise-ration (SNR). Operations at high field have the advantage of being able to enhance the SNR and gain a high spectral resolution. However, to realize the full benefits of high field parallel MRI, current radio frequency (RF) technology still needs improvements and many technical challenges need to be overcome, such as providing an optimized engineering solution which balances the RF magnetic field penetration, coil sensitivity, coupling/decoupling and B1 field homogeneity. Moreover, in some applications, such as in a small animal system, the limited space within small animal systems makes it difficult to incorporate both transmit and receive coils. In this thesis, these issues concerning transceive systems are explicitly optimized for RF performance through a number of proposed novel and effective hardware solutions. An inverse method for the design of phased-array is described and applied to the design of asymmetric, unshielded RF array coils. This method expands the geometry of arrays from conventional cylinders to part-spheres, cone and ellipsoid shapes that conform more closely to the anatomy under study. The research work developed and presented in this thesis has been applied to two specific RF coil applications, human breast imaging and small animal imaging. Firstly, for transceive RF breast coils design, a two-loop array coil has been designed and optimized to improve the homogeneity of the B1 field. Secondly, a hybrid structured design of the breast array coil has been proposed for the improvement of the overall RF field performance, especially the superior-inferior area of the breast, which is a problematic area for the conventional breast coil. Finally, a single-turn solenoidal transceive breast coil design for a bilateral breast MRI has been introduced. For small animal MRI in a high field application, three dedicated, shielded eight-element transceive volume-arrays for large rat MRI applications has been designed and two have been constructed. Several techniques and methods have been proposed and employed for the array design to help to improve the coil RF penetration depth and the signal reception capability, and also to minimize the mutual decoupling between adjacent coil elements. As an important accessory of the phase array system, a high power eight-channel T/R switch unit is also designed and fabricated. Both experimental and theoretical studies have demonstrated the technical potential of the proposed transceive phased-array technology for high-field human and small animal MRI applications.
KeywordPhased array
Inverse Design
Hybrid design
RF breast coil
RF small animal coil
Transceive
Parallel MRI
Additional Notespage number of colour page of PDF file p37, p39, p40, p41, p43, p47, p49, p51, p52, p53, p54, p56, p57, p58, p60, p61, p63, p64, p69, p71, p72, p76, p83, p84, p89, p94, p102, p116, p123, p129, p131, p132, p133, p134, p139, p140, p141, p142, p143, p151, p152, p153, p154, p155, p156, p161, p164, p165, p166, p167, p169, p170, p172, p173, p174, p175, p180, p181, p204, p205, p206, p207, p209, p210, p211, p212, p213, p214, p215, p216, p217, p231, p232, p234, p235, p236, p237, p238, p239, p240, p241, p246, p247

内容概要:本文详细探讨了基于樽海鞘算法(SSA)优化的极限学习机(ELM)在回归预测任务中的应用,并与传统的BP神经网络、广义回归神经网络(GRNN)以及未优化的ELM进行了性能对比。首先介绍了ELM的基本原理,即通过随机生成输入层与隐藏层之间的连接权重及阈值,仅需计算输出权重即可快速完成训练。接着阐述了SSA的工作机制,利用樽海鞘群体觅食行为优化ELM的输入权重和隐藏层阈值,从而提高模型性能。随后分别给出了BP、GRNN、ELM和SSA-ELM的具体实现代码,并通过波士顿房价数据集和其他工业数据集验证了各模型的表现。结果显示,SSA-ELM在预测精度方面显著优于其他三种方法,尽管其训练时间较长,但在实际应用中仍具有明显优势。 适合人群:对机器学习尤其是回归预测感兴趣的科研人员和技术开发者,特别是那些希望深入了解ELM及其优化方法的人。 使用场景及目标:适用于需要高效、高精度回归预测的应用场景,如金融建模、工业数据分析等。主要目标是提供一种更为有效的回归预测解决方案,尤其是在处理大规模数据集时能够保持较高的预测精度。 其他说明:文中提供了详细的代码示例和性能对比图表,帮助读者更好地理解和复现实验结果。同时提醒使用者注意SSA参数的选择对模型性能的影响,建议进行参数敏感性分析以获得最佳效果。
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值