In the last decade, MRI studies of human brain morphometry have been used to investigate a multitude of pathologies and drug-related effects in psychiatric research. The morphometric measures that differentiate patient populations or track longitudinal changes are often subtle and require a large number of subjects or repeated studies to detect and statistically model with significance. Cost, patient compilance, risks to the patients, and the rarity of certain diseases often limit traditional, clinical morphometric studies. These complications have motivated the analysis of mice and rats. Such animal studies are also very popular due to the animals'small size, their rapid life cycle, the wealth of knowledge about their behavior and development, as well as the maturity of the technology to manipulate their genetic information to induce disease. Brain morphometry models of mice and rats typically involve histological slides, behavioral data, genetic testing, and, increasingly, MRI scans. In particular, MRI scans are being employed as a hypothesis generation method for focused histological and molecular examinations in addition to brain morphometry and connectivity. While effective methods have been developed for automatically extracting brain morphometry from human MRI scans, few automated quantitative analysis methods exist for small animal MRI. For human MRI image analysis, our group is one of the leaders in the field for development and application, as we have developed and applied methods for bias correction, atlas-based tissue classification, structural segmentation, diffusion tensor analysis as well as localized shape analysis. In contrast to the human MRI brain imaging, the standard for murine MRI brain analysis is to manually outline brain features in each slice for a large number of scans. Such manual methods lack reproducibility and are very time-consuming. The lack of automated MRI analysis methods is the limiting factor in many animal studies. In this work we propose to continue our successful Phase I work to develop automatic, reliable, high-throughput MR image analysis methods for small animal, brain morphometry studies. These methods are available to investigators via an intuitive web- based interface for collecting, distributing, and automatic processing the imaging data in small animal studies. The web-based data sharing and processing system also supports the inspection of the ongoing processing and the examination of the computed results. This web-based processing system is generic in nature and can be extended to host and process human MRI data as well as data from other modalities and other applications. To demonstrate and evaluate the whole system, our collaborators will apply it to several studies of murine and rat brain morphometry currently conducted at UNC. The feedback collected from these studies will directly improve the usability of the proposed system. The proposed software will advance murine MRI studies of morphometry and connectivity for neuro-developmental, and neuro-degenerative psychiatry diseases. Our goal is that the analysis of MR images of entire brain studies will become the matter of a few mouse-clicks on a web-interface. Key personnel: Martin Styner, Neuro Image Research and Analysis Laboratory, Depts. Psychiatry and Computer Science Stephen Aylward, Julien Jomier, Kitware Inc.
High Throughput Web-based Image Analysis of Mouse Brain MR Imaging Studies Project Narrative The shapes of structures in the human brain, imaged using MRI, have been associated with a multitude of pathologies and drug-related effects in psychiatric research. However, cost, risks to the patients, and the rarity of certain diseases often limit the feasibility such studies. These complications have motivated the use of model organisms such as of mice and rats for such studies. We propose to develop automatic, reliable, high- throughput MR image analysis methods for small animal, brain studies. Additionally, we propose to develop an intuitive web-based interface for collecting, distributing, and processing the imaging data, as well as grouping it with intermediate results and final publications.
Showing the most recent 10 out of 18 publications
