The Biomedical Image Computing and Informatics Cluster (BICIC) will meet the rapidly growing needs of biomedical image computing research at Penn, and at the Center for Biomedical Image Computing in particular, and of the center's network of NIH-funded collaborating studies. Biomedical image computing faces several challenges. Increased algorithmic complexity demands computationally intensive computing and data mining of large collections of imaging, clinical and genetic data from growing patient populations is needed in order to discover biologically and clinically important relationships. These challenges underline the need for the advanced computing and storage facilities that the BICIC will provide. The proposed instrument represents an approximate 7-fold increase over the currently available resources, allowing both a more rapid execution of existing computerized analyses and providing the ability to explore methods that are currently infeasible with current equipment. The availability of this computing power in a single facility, instead of scattered resources of individual labs, will enable collaboration on algorithms, programming methods, datasets, and processing techniques that is not currently possible. The proposed server will provide a platform for developers and users of these sophisticated and demanding algorithms to push the envelope of biomedical image computing science to new levels. The proposed supercomputer will allow for high throughput analysis of scans, accelerating knowledge discovery and design of further analyses. The contribution of such a system to basic scientific research will be immense, as many scientific projects that are now infeasible, or which require computation measured in weeks or months, will produce results within minutes or days. The facility will encourage rapid development of complex image and connectomic analysis, pattern recognition, and data mining algorithms, often working on high- dimensional multi-parametric data, thereby allowing us to maximize the amount and accuracy of information gathered from biomedical images. Data mining of large databases and of complex data will expose new relationships between genotypes and phenotypes, and will potentially reveal subtle characteristics of certain pathologies that have clinical values. It will also aid Penn's strong focus on translational research in the field of medical imaging, which is currently limited by the lack of a sufficiently powerful computer system to facilitate the demanding imaging studies. The basic and clinical research that the proposed computational server will enable is expected to have a very significant clinical impact, underlining the importance of the project.
Computational science is playing an increasing role in biomedical applications, particularly in the analysis of medical images. The proposal requests computer resources to meet computational and storage needs that will aid the advanced development of biomedical image computing methods being researched at the University of Pennsylvania and their application to the existing body of medical image data. This advanced computational resource will facilitate the development and application of advanced computational imaging algorithms that is expected to pave the way for research with high clinical impact.
