Development and Dissemination of Robust Brain MRI Measurement Tools Abstract: Summary. Neuroimaging provides a safe, non-invasive measurement of the whole brain, and has enabled large clinical and research studies for brain development, aging, and disorders. However, many disorders, i.e., major neurodegenerative and neuropsychiatric disorders, cause complex spatiotemporal patterns of brain alteration, which are often difficult to identify visually and compare over time. To address this critical issu, in the renewal phase of this project, we will continue to work with GE Research to develop and disseminate a software package for brain measurement, comparison, and diagnosis. The new tools include 1) a novel tree-based registration and multi-atlases-based segmentation method for precise measurement of brain alteration patterns, and 2) novel pattern classification and regression methods for early detection and longitudinal monitoring of brain disorders.
Aims. Currently, most existing atlas-based labeling methods simply warp each atlas independently to the individual brain for multi-atlases-based structural labeling. This could lead to 1) inaccurate labeling due to possible large registration error when the atlases are very different from the target individual brain, and 2) inconsistent labeling of the same brain structure across different individuals due to independent labeling of each individual brain. The first goal of this project is hence to develop a novel tree-based registration and multi-atlases -based segmentation method for simultaneous registration and joint labeling of all individual brains by concurrent consideration of all atlases. With measurements of brain structures and their alteration patterns, univariate analysis methods are often used to understand how the disease affects brain structure and function at a group level. Although this can lead to better understanding of neurological pathology of brain disorders, more sophisticated image analysis methods are urgently needed for quantitative assessment and early diagnosis of brain abnormality at an individual level. Thus, the second goal of this project is to develop various novel machine learning methods for early diagnosis of brain disorders and better quantification of brain abnormality at an individual level. Specifically, we will take Alzheimer's disease (AD), which is the most common form of dementia, as an example for demonstrating the performance of our proposed methods in early diagnosis of AD, as well as in prediction of long-term outcomes of individuals with mild cognitive impairment (MCI). The last goal of this project is to build, for ou developed methods, the respective software modules for the 3D Slicer (a free open-source software package with a flexible modular platform for medical image analysis and visualization, www.slicer.org/), to promote the potential clinical applications by using tools in 3D Slicer for preprocessing of patient data and our tools for diagnosis. Again, this software development work will be performed in collaboration with our current collaborator, GE Research, which is a part of the engineering core of the National Alliance for Medical Image Computing (NA-MIC) that is focused on developing 3D Slicer. Both source code and pre-compiled programs will be made freely available. Applications. These methods can find their applications in diverse fields, i.e., quantifying brain abnormality of neurological diseases (i.e., AD and schizophrenia), measuring effects of different pharmacological interventions on the brain, and finding associations between structural and cognitive function variables.
Description of Project This project aims to develop a novel method for accurate measurement of brain structure and function by groupwise registration and joint labeling of all individual images via multiple manual-labeled atlases. Moreover, several novel tools for brain abnormality measurement will also be developed for early detection and progression monitoring of brain disorders using both multimodal imaging and non-imaging data. By successful development of these brain measurement tools, the respective software modules will be developed and further incorporated into 3D Slicer via collaboration with GE Research, which is a part of the engineering core of the National Alliance for Medical Image Computing (NA-MIC). Both source code and pre-compiled programs will be made freely available.
|Yin, Qingbo; Hung, Sheng-Che; Wang, Li et al. (2017) Associations between Tumor Vascularity, Vascular Endothelial Growth Factor Expression and PET/MRI Radiomic Signatures in Primary Clear-Cell-Renal-Cell-Carcinoma: Proof-of-Concept Study. Sci Rep 7:43356|
|Hu, Shunbo; Wei, Lifang; Gao, Yaozong et al. (2017) Learning-based deformable image registration for infant MR images in the first year of life. Med Phys 44:158-170|
|Rekik, Islem; Li, Gang; Yap, Pew-Thian et al. (2017) Joint prediction of longitudinal development of cortical surfaces and white matter fibers from neonatal MRI. Neuroimage 152:411-424|
|Chen, Xiaobo; Zhang, Han; Lee, Seong-Whan et al. (2017) Hierarchical High-Order Functional Connectivity Networks and Selective Feature Fusion for MCI Classification. Neuroinformatics 15:271-284|
|Chen, Xiaobo; Zhang, Han; Zhang, Lichi et al. (2017) Extraction of dynamic functional connectivity from brain grey matter and white matter for MCI classification. Hum Brain Mapp 38:5019-5034|
|An, Le; Adeli, Ehsan; Liu, Mingxia et al. (2017) A Hierarchical Feature and Sample Selection Framework and Its Application for Alzheimer's Disease Diagnosis. Sci Rep 7:45269|
|Zu, Chen; Wang, Zhengxia; Zhang, Daoqiang et al. (2017) Robust multi-atlas label propagation by deep sparse representation. Pattern Recognit 63:511-517|
|Shen, Dinggang; Wu, Guorong; Suk, Heung-Il (2017) Deep Learning in Medical Image Analysis. Annu Rev Biomed Eng 19:221-248|
|Meng, Yu; Li, Gang; Rekik, Islem et al. (2017) Can we predict subject-specific dynamic cortical thickness maps during infancy from birth? Hum Brain Mapp 38:2865-2874|
|Xiang, Lei; Qiao, Yu; Nie, Dong et al. (2017) Deep Auto-context Convolutional Neural Networks for Standard-Dose PET Image Estimation from Low-Dose PET/MRI. Neurocomputing 267:406-416|
Showing the most recent 10 out of 278 publications