Large-scale analysis of neuroimaging data on commercial clouds with content-aware resource allocation strategies

Abstract

The combined use of mice that have genetic mutations (transgenic mouse models) of human pathology and advanced neuroimaging methods (such as MRI) has the potential to radically change how we approach disease understanding, diagnosis and treatment. Morphological changes occurring in the brain of transgenic animals as a result of the interaction between environment and genotype, can be assessed using advanced image analysis methods, an effort described as "mouse brain phenotyping". However, the computational methods involved in the analysis of high-resolution brain images are demanding. While running such analysis on local clusters is possible, not all users have access to such infrastructure and even for those that do, having additional computational capacity can be beneficial (e.g., to meet sudden high throughput demands). In this paper we use a commercial cloud platform for brain neuroimaging and analysis. We achieve a registration-based multi-atlas, multi-template anatomical segmentation, normally a lengthy in time effort, within a few hours. Naturally, performing such analyses on the cloud entails a monetary cost, and it is worthwhile identifying strategies that can allocate resources intelligently. In our context a critical aspect is the identification of how long each job will take. We propose a method that estimates the complexity of an image processing task, a registration, using statistical moments and shape descriptors of the image content. We use this information to learn and predict the completion time of a registration. The proposed approach is easy to deploy, and could serve as an alternative for laboratories that may require instant access to large high performance computing infrastructures. To facilitate adoption from the community we release publicly the source code.

Semi-automated registration-based anatomical labelling, voxel based morphometry and cortical thickness mapping of the mouse brain

“…The methodological workflow presented in this work was designed to facilitate the implementation of fine-grained morphoanatomical mapping tools by non-expert users, and promote forward and back translation of MRI preclinical and clinical research evidence. We also point out that a preliminary account on the implementation of these procedures in parallel computing cloud environment has been recently reported (Minervini et al, 2014), a strategy that can streamline and accelerate image processing time by exploiting large high-performancecomputing infrastructures.…”

“…To simplify and streamline operations, we based image processing mainly on ANTs (Avants et al, 2009), a flexible and powerful open source toolkit freely available to the scientific community. Importantly, our approach has been recently applied by our research group to map fine-grain brain anatomy alterations in different mutant mouse lines (Dodero et al, 2013;Lassi et al, 2015;Minervini et al, 2014;Sannino et al, 2014; and to describe large-scale networks of anatomical covariance between gray matter regions in wild-type mice (Pagani et al, 2016), with excellent agreement between MRI and manual morphometric measurements (Sannino et al, 2014), exhibiting corresponding morphoanatomical features in mice and reference clinical populations (Cutuli et al, 2016;Tucci et al, 2014). Below, we provide a detailed description of our procedural workflow and show its capabilities by describing its application to quantify morphological alterations in sociallyimpaired BTBR T+Itpr3tf/J mice with respect to normo social C57BL/6J controls (Dodero et al, 2013;Squillace et al, 2014), a comparison that has been recently described by our research group (Dodero et al, 2015) and others (Ellegood et al, 2013), thus permitting an empirical crosslaboratory assessment of the validity of our findings.…”

Science in the cloud (SIC): A use case in MRI connectomics

“…All of the code for this project is open-source and resides in a Github repository 14 . To test the pipeline with different sets of parameters, it can be cloned and the source code can be modified directly.…”

“…Leveraging the NeuroDebian platform, NITRC has encouraged a transition to the cloud by releasing an Amazon Machine Image (AMI) preloaded with commonly used packages. In parallel, many groups have strived to breach the frontier through such efforts as developing sophisticated resource estimation-based deployment strategies [ 14 ], and these have shown the great potential for a cloud-based approach to neuroimaging [ 15 ]. Each of these projects has made valuable contributions to the progress towards accessibility and portability of neuroscience research.…”

Running Neuroimaging Applications on Amazon Web Services: How, When, and at What Cost?

“…Performance may also vary because of characteristics that are not clearly visible to the user, such as the hardware heterogeneity that underlies virtual instances, scheduling policies used by the cloud provider, and the effects of running multiple virtual instances together with other users on the same hardware (otherwise known as multitenancy). Finally, performance can be unpredictable because of the characteristics of the application itself ( Minervini et al, 2015 ). The variability of execution time and its impact on cost must also be considered.…”