High Throughput Landmark Based Image Registration Using Cloud Computing

Cloud computing provides an elastic, as-a-Service resource abstraction with a usage-based payment model, and is emerging as an attracting computing paradigm. This is especially true for computing intensive application workflows with high-throughput requirements, such as those involving large-scale image datasets. In this paper, we investigate the usage of cloud computing to support a medical image registration workflow using the CometCloud autonomic cloud computing framework. CometCloud supports the execution on workflows on heterogeneous, dynamically federated (private and public) clouds, where the federation is driven by application requirements and user objectives. Specifically, in this investigation, we use objectives such as workflow execution deadlines and budget constraints to provision the appropriate type and number of resources, which are the federated with local private resources to provide additional resources to satisfy the objectives. We experimentally tested this cloud-based workflow on a large dataset containing two sets of consecutive digitized breast tissue specimens. This cloud computing based image registration workflow shows promising results in its ability to meet high-throughput, deadline driven requirements.