Development, Optimization and Implementation of New Algorithms for Improving the Resolution of EM, Acoustic and fMRI Reconstructions

许多人熟悉通过磁共振(MR)成像获得的人体内部图像。与摄影图像不同，磁共振图像不是直接作为图像捕获的，而是必须使用离散傅立叶变换(DFT)算法从体内氢核激发的信号中计算出来。有人说“用来分析任何一天的信号的DFT算法比任何其他算法都多”。这意味着“有很多研究机会可以用来理解和解决DFT的局限性。”一个基本的问题是，必须在空间分辨率、所捕获的MR图像的质量和时间分辨率之间做出妥协。时间分辨率是跟踪变化的能力，例如对运动的心脏进行成像。错误的信息会出现在图像中，这会混淆诊断。 我的团队最近将我们自己的用于提高空间分辨率的约束重建与提出的稀疏采样方法相结合，这些方法通过收集更少的数据来故意节省时间。一些意想不到的初步结果被发现，图像的一些部分提高了质量，而另一些部分则变得更加扭曲。我们正在开发一种新的后处理重建过程，能够对MR信号进行数学处理，以产生具有满足诊断需要的质量的图像。这需要将许多质量不佳的图像的最佳部分结合在一起--类似于最近的想法，即用软件将许多平坦、廉价、低分辨率的镜头组合在一起，取代手机中笨重(厚重)、昂贵的高分辨率相机镜头。 我们已经确定，我们的新方法能够提高各种信号和图像重建算法的能力：(1)使用稀疏的声传感器阵列来收集强度和相位信息，导致工业和社区噪声定位，以帮助我们的社区声学噪声最近受到相当大的媒体关注，(2)改进低分辨率的fMRI图像，为诊断视神经炎(ON)提供新的标记。ON通常是导致多发性硬化症的初步阶段；不幸的是在诸如加拿大之类的北部国家很常见，以及(3)改进工业实时实验室工作台嵌入式系统在表征使用稀疏2D霍尔效应传感器阵列测量的磁近场效应方面的分辨率， 在实验室开发这些算法是一回事，但我们必须能够将这些算法应用于实际的工业、医疗和社区情况。我们一直在跟踪最近成功的业务(企业)“面向敏捷的软件过程”进入可靠的嵌入式系统代码的动向。由于缺乏文献，我们建议使用我们提出的新的重建算法来进行系统的调查，以在一个研究领域使用敏捷思想；一个潜在的广泛受众的领域。

Many people are familiar with the internal body images available through magnetic resonance (MR) imaging. Unlike photographic images, MR images are not captured directly as an image but must be calculated using a discrete Fourier transform (DFT) algorithm from signals excited from the hydrogen nuclei in the body. It has been said that “There are more DFT algorithms used to analyze signals on any given day than any other algorithm”. This implies “There are many research opportunities available to understand and solve the limitations of DFTs." A basic problem is that a compromise must be made between spatial resolution, the quality of the MR image captured, and time resolution, the ability to track changes such as imaging the moving heart. False information appears in the image which confuses diagnosis. My team has recently combined our own constrained reconstruction for improving spatial resolution with proposed sparse sampling approaches that deliberately time save by gathering less data. Some unexpected preliminary results were found, with some parts of the image improving quality while others became more distorted. We are developing a new post-processing reconstruction process capable of mathematically manipulating the MR signals to produce an image with all parts having the quality to meet diagnosis needs. This requires combining the best parts of many poor quality images -- similar to recent ideas of replacing the bulky (thick), expensive, high-resolution camera lens in mobile phones with software that combines images from many flat, inexpensive, low-resolution lenses. We have identified that our new approach has the capability of improving the capability of a wide range of signal and image reconstruction algorithms: (1) using a sparse array of acoustic sensors to gather intensity and phase information leading to industrial and community noise localization to help with our Community Acoustic Noise which has recently received considerable media attention, (2) improving the low-resolution fMRI images to provide new markers to diagnose Optical Neuritis (ON). ON is often a preliminary stage leading to Multiple Sclerosis; unfortunately common in northern countries such as Canada and (3) improving an industrial real-time lab bench embedded system’s resolution in characterizing magnetic near-field effects measured using a sparse 2D hall-effect sensor array, Developing these algorithms in the laboratory is one thing, but we must be able to apply these algorithms in practical industrial, medical and community situations. We have been following up on the recent movement of successful business (enterprise) “Agile oriented software processes” into reliable embedded system code. The lack of literature has led us to suggest we use our proposed new reconstruction algorithms to undertake systematic investigations into using Agile ideas in a research field; an area with a potential wide audience.