Author:

Keywords:

PSI_MIC

Abstract:

Early diagnosis, treatment follow up and devising new therapeutic strategies for neurodegenerative diseases, have both clinical and socio-economic impact. Investigating the experimental animal models of neurodegeneration can provide the required insight. Conventional investigation of genetically engineered animals (e.g. rats, mice) are based on histological analysis, which requires a large number of animals to be sacrificed. In contrast, in vivo imaging allows to repeatedly scan the same animal non-invasively. The dedicated state-of-art small-animal imaging systems (e.g. microMRI, microPET, etc.), allows macroscopic and microscopic assessment of structural changes, as well as those of metabolic and cellular processes. The potential of small animal models and their imaging can be fully harnessed, with objective analysis of multi-temporal and multi-modal images. In this research, we focused on the objective assessment of small animal brain images (primarily MRI) in the context of neuroimaging applications. Range of image artifacts and challenges confound tsunami of real world data generated by small animal imaging centers (8000-10000 images/year). This also makes conventional manual assessment tedious, time consuming and error-prone. Image artifacts such as MR non uniformity , pose variations , inter-scan intensity variations, etc. that affected the quality and quantification of our pre-clinical images, were first documented. A series of (semi-) automated image analysis methods for intensity normalization , spatial normalization and image segmentation were implemented to counter the confounding factors. Spatial and intensity normalization schemes enabled comparison of images from the same animal acquired at different time points, either in vivo or ex vivo , as well as joint interpretation of multi-modal image data (microMRI, microCT, microPET). Segmentation methods, together with anatomical template of the rodent brain allowed same volume-of-interest (e.g. ventricles, putamen, olfactory bulb, blood vessels) to be automatically identified across all images. Qualitative and quantitative evaluation of a large number of heterogeneous real-world images, revealed the inter-dependency between the above methods. Accordingly, a systematic workflow -small animal image analysis pipeline for assessment of brain MR images was established. The sequential integration was made comprehensive by including methods for format conversion, visualization as well as auxiliary tools for interpretation of results. The standardized workflow required minimum or no parameter tuning for each of the methods. Three different applications were considered to demonstrate the benefits of objective image quantification in small animal imaging: (1) assessment of morphological phenotyping, (2) evaluation of potential MRI reporters for stem cell labeling and (3) the use of vasculature information during stereotactic neurosurgery planning. Key results from these applications demonstrated the need and benefit for a more efficient and effective use of imaged data as it can expose subtle differences (e.g. morphological phenotyping) or changes over time (e.g. MRI cell labeling), as well as confounding effects in the experimental setup (e.g. injury to blood vessels during neurosurgery). We conclude that the image analysis workflow is generic and readily applicable for a wider scope of quantification needs of small animal neuroimaging applications, in life science research. Capitalizing on the combination of non-invasive imaging and objective assessment, the effective, efficient and ethical use of animal data were illustrated, which in turn has the potential to reduce the number of animals required for laboratory experiments.