Algorithm Development for Magnetic Resonance Images (MRI)

医学图像的视觉分析是主观的（依赖于图像质量）、定性的、容易出错的，并且是低效的，这最终影响诊断准确性和进行大型研究的能力。图像分析技术提供了更好的选择，因为它们以定量、客观、有效和可再现的方式自动测量图像属性。该计划的目的是推进最先进的图像分析算法，这些算法仅在液体衰减反转恢复（FLAIR）磁共振图像（MRI）上运行，以一种新颖有效的方式量化神经系统疾病。 尽管FLAIR MRI现在是用于神经退行性疾病检测和分析的优选模式，但仅存在少数图像处理方法，并且非常需要新颖的图像处理算法。本研究提案的目标围绕开发1）FLAIR MRI中神经解剖学和病理学的分割算法，以便从组织中提取特征，如体积、纹理和形状，以及2）FLAIR MRI与图谱的配准，以便在空间标准化空间中分析脑图像，研究纵向数据中的疾病进展（即病变体积的变化），并且还识别重要的解剖标志。考虑到神经系统疾病的患病率，人口老龄化以及加拿大医疗保健系统的相关经济负担，这样的计划有可能对许多加拿大人产生积极影响。

The visual analysis of medical images is subjective (observer-dependent), qualitative, error prone, and is inefficient, which ultimately effects diagnostic accuracy and the ability to conduct large research studies. Image analysis techniques offer a better alternative since they automatically measure image properties in a quantitative, objective, efficient, and reproducible manner. The purpose of this program is to advance the state-of-the-art in image analysis algorithms that operate only on Fluid-Attenuation Inversion Recovery (FLAIR) Magnetic Resonance Images (MRI) to quantify neurological disease in a novel and efficient way. Although FLAIR MRI are now the preferred modality for neurodegenerative disease detection and analysis, only few image processing methods exist and novel image processing algorithms are much needed. The objectives of this research proposal surround the development of 1) segmentation algorithms for neuroanatomy and pathology in FLAIR MRI in order to extract features from tissues such as volume, texture and shape, and 2) registration of FLAIR MRI to atlases in order to analyze brain images in a spatially normalized space, to study progression of disease in longitudinal data (i.e. change in lesion volume), and also to identify important anatomical landmarks. Considering the neurological disease prevalence, aging demographics, as well as associated economic burden on the Canadian healthcare system, such a program has the potential to positively impact many Canadians.