Medical image processing can be time-consuming, and accelerating the imaging will enable doctors to make more rapid diagnoses. To greatly increase the speed of processing, Dr. Bradley Erickson, a radiologist from Mayo Clinic in Rochester, Minn., and engineers from IBM Corp. in Armonk, N.Y., have integrated a parallel processor with software that contains an image processing algorithm. In tests, they determined that the processor and software can hasten the process of aligning magnetic resonance and CT images. Erickson said that MRI and CT scanners capture slices perpendicular to the imaging table, but patients tend to move during imaging, resulting in slices at different angles. “The algorithm finds the angle and recuts the images so they match,” he explained. “When two images are aligned, you maximize the ability of the intensity from one image to predict the intensity of the same location for the other image.”Because the algorithm corrects for misalignment, it improves imaging accuracy and speed. He said that this is especially important for patients — such as those with cancer — who undergo regular imaging to monitor ongoing treatment.During development, the researchers integrated the software with IBM’s “cell blade” processor, which comprises eight parallel processors to reduce processing time. Because medical imaging produces large amounts of data, they stored blocks of the existing data to make way for the new data as it was being recorded, which they called an “image stripe.” They processed 98 sets of MRI and CT images obtained from various Siemens and GE Healthcare scanners. Using a conventional processor, image registration took approximately 7 h. In contrast, the cell blade processor registered the images in only 516 s, or ~50 times faster. No image registration lasted more than 20 s, and most took less than 1 s. The researchers presented their results at the IEEE International Symposium on Biomedical Imaging in April. Erickson said that no one has ever achieved subsecond processing speeds for this type of imaging, and that he believes that the processor can be integrated into any three-dimensional medical imaging system to reach the same rapid speeds. He said the group is deciding how best to fit the processor into medical systems.