High-performance medical imaging informatics.

The topic of High Performance Medical Image Computing for Image-Assisted Clinical Intervention and Decision-Making is central for imaging informatics, since the amount and complexity of imaging data continues to grow as new modalities are added to the radiological repertoire in an increasingly distributed (and often cloud-computing) environment. This top ic was the subject of the 2010 High Performance – Medical Image Computing and Computer Assisted Intervention (HPMICCAI) Workshop on High Performance Medical Imaging, which we organized with a number of colleagues, and from which two papers are included in this issue of Methods of Information in Medicine as part of a Focus Theme. Recent advances in acquisition, sensing and automated technology have generated an explosion of high resolution and high frequency medical image and video data. However, as the acquisition of massive data from the new modalities increases, so does the need for more efficient ways of processing the data in both clinical practice and clinical research environments. The effective and efficient interpretation of suites of complex medical images depends on how well and how rapidly they can be analyzed, and how effectively summarized. Prolif eration of high resolution multimodality medical images and videos presents considerable challenges to both clinical practitioners and IT specialists. The value of the image datasets for time-critical clinical applications such as image-assisted surgery, cancer therapeutics, clinical decision-making, and multimedia data fusion and analytics, is undisputed. Conventional serial computation is inadequate and inefficient for handling a high volume of images of different modalities, and can significantly affect quality of care, by delaying diagnoses and treatments. High-performance (HP) computing holds the key for unlocking the full potential of medical imaging. More specifically, serial computation is inadequate for handling large amounts of medical image data because it typically involves computation times in the order of hours and often days and frequently must resort to sub-sampling of the data. Recently, exceedingly powerful computer hardware and optimized image processing software can, for the first time, allow highvolume image data processing and mani pulation to become clinically feasible on a routine basis in real-time. High-performance computer-cluster, multi-processor, multi-core and many-core technology with high-volume throughput and vector processing capabilities promise to reveal the clinically useful information contained in medical imaging data and have a strong impact in transferring medical imaging research from the lab into clinical practice. Increasingly multi-disciplinary research efforts have sprung up to develop innovative designs, techniques and algorithms that ex-