Computational Tools for the Analysis of MRI Images in Type-1 Diabetes

 描述(申请人提供)：该项目旨在通过开发新的计算图像分析算法，根据磁共振成像(MRI)数据量化这些变化，从而表征与1型糖尿病(T1D)相关的胰岛炎症(称为胰岛素炎)和体积减少的局部变化。仅在美国，就有多达300万人可能患有T1D，每天有80个新诊断病例，每年花费近150亿美元(来源：JDRF)。在器官水平上了解细胞自身免疫破坏的机制对于制定新的早期诊断标准、有效的治疗策略和预防性治疗具有重要意义。许多自身免疫过程的临床隐蔽性，加上很难获得朗格汉斯内分泌胰岛的位置，减缓了对T1D病因和进展的了解。然而，MRI通过使用磁性纳米颗粒(MNP)试剂，允许对胰腺解剖结构(如体积)进行非侵入性的局部测量，并通过使用磁性纳米颗粒(MNP)试剂来减轻这一问题，从而使横断面和纵向T1D成像研究成为可能。为此，需要两个或多个图像的胰腺区域之间的准确对应，以便计算1)来自注射MNP-MR图像的胰岛素炎，2)胰岛素炎随时间的进展，以及3)胰腺体积随时间的局部变化，以及4)比较受试者之间的所有这些量。这种逐点对应是通过图像配准(对准)来提供的。这项建议旨在以我在脑图像分析方面的背景为基础，开发新的图像配准(对齐)工具，以准确计算从不同受试者在不同时间获取的胰腺图像之间的点状对应关系，并随后将其用于横断面和纵向胰腺成像，通过局部跟踪临床和/或隐匿性T1D患者的长期炎症和体积变化，开发新的生物标记物。具体地说，我们提出了一种固有对称的准体积保持(QVP)非刚性胰腺图像配准算法，与文献中已有的算法不同，该算法通过定义统一的目标函数来避免区域偏差和伴随的误差。此外，使用一种新的改进的分组配准算法，通过构建健康和T1D队列的无偏统计胰腺地图集来衡量组内差异和组内变异性。我的长期职业目标是在一所顶尖的学术机构建立和指导一个跨学科的研究项目，该项目将专注于开发处理生物医学图像的创造性方法和创新的计算工具，以促进对医学图像和临床数据之间关系的调查，并改善患者的诊断和结果。我在K01获奖期间的主要目标是成为T1D方面的专家，除了腹部-特别是胰腺-MR采集和图像分析，并通过转化我之前在脑图像重建和分析方面获得的技能来推动这一领域的发展。为了实现这一目标，我在三个重要领域需要额外的培训、指导和经验：1)糖尿病，2)腹部造影剂造影剂，3)先进的研究设计和生物统计学。我建议通过直接指导、授课课程、模块课程、系列研讨会和科学会议来获得这种培训。提议的项目将构成我的独立计算腹部成像研究计划的基础，该计划将把糖尿病作为其临床重点的核心。

 DESCRIPTION (provided by applicant): This project aims to characterize the local changes in pancreatic islet inflammation (referred to as insulitis) and volume loss associated with type-1 diabetes (T1D) by developing novel computational image analysis algorithms that quantify such changes from magnetic resonance imaging (MRI) data. In the United States alone, as many as three million people may have T1D, with 80 new cases diagnosed every day, costing almost $15B annually (source: JDRF). Understanding the mechanisms of autoimmune destruction of ß cells at the organ level is important for developing new early diagnostic criteria and effective treatment strategies and preventative therapies. Clinical occultness of much of the autoimmune process, along with the difficult access to the location of the endocrine islets of Langerhans have slowed progress in understanding the etiology and progression of T1D. However, MRI alleviates this by permitting noninvasive, local measurement of pancreatic anatomy (such as the volume), in addition to insulitis via the use of magnetic nanoparticle (MNP) agents, making cross-sectional and longitudinal T1D imaging studies feasible. To that end, accurate correspondence among pancreatic regions of two or more images are required in order to compute 1) insulitis from pre/post- infusion MNP-MR images, 2) the progress of insulitis over time, and 3) the local change in pancreatic volume over time, in addition to 4) comparing all of these quantities across subjects. Such a point-wise correspondence is provided by image registration (alignment). This proposal aims to build on my background in brain image analysis and develop novel image registration (alignment) tools to accurately compute point- wise correspondence between pancreas images acquired from different subjects at different times, and subsequently use them in cross-sectional and longitudinal pancreatic imaging to develop new biomarkers, by locally tracking long-term inflammatory and volume changes in individuals with clinical and/or occult T1D. Specifically, we propose to develop an inherently-symmetric quasi-volume-preserving (QVP) non-rigid image registration algorithm for the pancreas, which, in contrast to the existing algorithms in the literature, avoids regional biases and the concomitant errors by defining a uniform objective function. Furthermore, the intergroup differences and intragroup variability are measured by constructing unbiased statistical pancreatic atlases of healthy and T1D cohorts, using a novel, improved group-wise registration algorithm. My long-term career goal is to establish and direct an inter-disciplinary research program at a top-notch academic institution, which will focus on developing creative approaches and innovative computational tools for processing biomedical images, in order to facilitate the investigation of the relationship between medical images and clinical data, and improve patient diagnosis and outcomes. My main objective for the K01 award period is to become an expert in T1D, in addition to abdominal - and especially pancreatic - MR acquisition and image analysis, and to advance this field by translating the skills I had previously acquired in brain image reconstruction and analysis. To achieve this goal, there are three important areas where I require additional training, mentoring, and experience: 1) diabetes, 2) abdominal imaging with contrast agents, and 3) advanced study design and biostatistics. I propose to acquire this training through direct mentoring, didactic coursework, modular courses, seminar series, and scientific meetings. The proposed project will form the foundation of my independent computational abdominal imaging research program, which will have diabetes at the core of its clinical focus.