Wrangling Big Data Radiology with Medical Image Processing and High-Performance Computing
Structural localization of anatomy provides an essential framework to develop image-derived biomarkers, perform image quantification, assess longitudinal changes within a patient, and understand group differences through imaging. The image processing technologies of multi-atlas labeling, shape models, and machine learning have advanced to the point that we can reasonably identify almost any structure within typical MRI or CT scans. Multi-atlas labeling techniques now enable ready generalization of information from example images to derive automated image segmentation procedures. We will discuss advances from the Medical-image Analysis and Statistical Interpretation (MASI) lab in label fusion for neurologic, ophthalmologic, and abdominal imaging and present opportunities for big data applications through collaborative research and clinical data reuse.