The human brain undergoes a dynamic phase of development with rapid structural and functional growth in the first year of life. Insight into thi critical period of development is of paramount importance for understanding the neurodevelopmental origins of psychiatric illness, since brain alterations that are associated with psychosis and other major psychiatric illnesses often occur early during fetal or neonatal life. The recent availability of infant neuroimaging data is making increasingly feasible the precise characterization of development patterns in this period of time. However, computational tools that are dedicated to this purpose are still rare due to the following challenges: (1) Infant scans suffer from significantly lower spatial resolution due to the smaller brain size; (2) Limited by scn time, the achievable signal-to-noise ratio for diffusion-weighted images is typically low; (3) The rapid myelination process results in significant variation of image contrast across different brain regions, which can easily confuse existing computational methods; (4) Techniques developed for adult brain analysis are not directly transferable to infants. This project shoulders the challenging task of overcoming important technological hurdles in creating high- precision computational tools that will automate the quantification of brain development in the first year of life.
In Aim 1, we will create a 4D multimodality-guided, level-set-based framework for simultaneous segmentation and registration of serial brain scans acquired from birth to one year of age. This will allow low-contrast images (e.g., the isointense 3- and 6-month scans) to be segmented more effectively by borrowing multimodality information from early time-point (2-week) and/or later time-point (1-year) scans.
In Aim 2, we will create a 4D cortical surface reconstruction method for consistent surface reconstruction across different time points. This will help alleviate the imprecision stemming from structural ambiguities in the surface reconstruction process due to low image contrast.
In Aim 3, we will create a clustering-based hierarchically organized registration framework that will harness the manifold of anatomical variation of the image population for effective registration of infant brains. This will aid effectve registration of images with large structural differences to a common space for population-based early brain development studies.
In Aim 4, we will create super-resolution atlases for infant brains at each time point by using a novel patch-based sparse representation technique. These atlases, when used as templates for brain registration, will lead to significant performance improvement due to their significantly improved structural clarity. All created tools and super-resolution atlases will be integrated into a dedicated infant-brain-analysis software package and made freely available to the research community.
The computational tools created in this project are directly relevant to public health since they will lead to greater understanding of brain development and to greater capability in identifying neurodevelopmental origins of psychiatric illness. PUBLIC HEALTH RELEVANCE: This project aims at addressing a challenging problem of how to accurately segment serial infant brain images acquired in the first year of life (such as at an approximate interval of 3 months from birth to 1- year-old) and further quantify their longitudinal changes for early brain development study. The successful development of the proposed infant-brain-analysis tools will help understand early brain development and the future studies of neurodevelopmental disorders in the first years of life, and will also help fill up the knowledge gap of early brain development in this period. The final developed infant-brain-analysis algorithms will be made freely available to the research community via NITRC (Neuroimaging Informatics Tools and Resources Clearing house).
| Li, Guannan; Liu, Mingxia; Sun, Quansen et al. (2018) Early Diagnosis of Autism Disease by Multi-channel CNNs. Mach Learn Med Imaging 11046:303-309 |
| Meng, Yu; Li, Gang; Wang, Li et al. (2018) Discovering cortical sulcal folding patterns in neonates using large-scale dataset. Hum Brain Mapp : |
| Liu, Mingxia; Gao, Yue; Yap, Pew-Thian et al. (2018) Multi-Hypergraph Learning for Incomplete Multimodality Data. IEEE J Biomed Health Inform 22:1197-1208 |
| Rekik, Islem; Li, Gang; Lin, Weili et al. (2018) ESTIMATION OF SHAPE AND GROWTH BRAIN NETWORK ATLASES FOR CONNECTOMIC BRAIN MAPPING IN DEVELOPING INFANTS. Proc IEEE Int Symp Biomed Imaging 2018:985-989 |
| Wu, Zhengwang; Li, Gang; Wang, Li et al. (2018) CONSTRUCTION OF SPATIOTEMPORAL NEONATAL CORTICAL SURFACE ATLASES USING A LARGE-SCALE DATASET. Proc IEEE Int Symp Biomed Imaging 2018:1056-1059 |
| Zhang, Yongqin; Shi, Feng; Cheng, Jian et al. (2018) Longitudinally Guided Super-Resolution of Neonatal Brain Magnetic Resonance Images. IEEE Trans Cybern : |
| Wang, Fan; Lian, Chunfeng; Xia, Jing et al. (2018) CONSTRUCTION OF SPATIOTEMPORAL INFANT CORTICAL SURFACE ATLAS OF RHESUS MACAQUE. Proc IEEE Int Symp Biomed Imaging 2018:704-707 |
| Lian, Chunfeng; Liu, Mingxia; Zhang, Jun et al. (2018) Automatic Segmentation of 3D Perivascular Spaces in 7T MR Images Using Multi-Channel Fully Convolutional Network. Proc Int Soc Magn Reson Med Sci Meet Exhib Int Soc Magn Reson M 2018: |
| Liu, Mingxia; Zhang, Jun; Adeli, Ehsan et al. (2018) Landmark-based deep multi-instance learning for brain disease diagnosis. Med Image Anal 43:157-168 |
| He, Ping; Xu, Xiaohua; Zhang, Han et al. (2018) Spatiotemporal Analysis of Developing Brain Networks. Front Neuroinform 12:48 |
Showing the most recent 10 out of 205 publications
