Digital Imaging Processing Core

University Of Michigan At Ann Arbor, Ann Arbor MI

Investigators

Linked publications & trials

Abstract

The goals of the Digital Image Processing Core are to continue provide analytic tools for all projects. These unique image analysis tools are necessary for unbiased, accurate, quantitative analysis of temporal volumetric changes in single and multimodal MRI data sets of lab animals and human patients. Currently all MRI acquisitions independent of weightings for all interval exams of both animal and human studies are fully automatically registered to a "reference" data set in the first exam typically using a rotate- translate geometry model;this is a multimodality registration problem using existing software tools. Typically a T1-weighted, post-Gad sequence, is used as the reference for registering all successive interval exam sets. B1-field corrections have been found to be unnecessary thus far. In cases where the acquisitions involve high gradient fields such as those used for diffusion or perfusion weighted imaging, registrations are accomplished using a full affine model to support correction of shears caused by the gradients. Currently all human scan data including both original acquisitions and registered datasets, as well as animal backups, are stored on the Core's disk system. In the future all animal scans will be likewise stored on the Core's disks which can be accessed via TCP/IP or SAMBA protocols over 100 Mb ethernet. Of course the Core is responsible for maintaining data integrity and backup. The core will also develop the ability to track the positions of voxels within treated lesions across interval exams using high degree of freedom warpings. This facility for both animal and human imaging will support further investigation of the role of the apparent diffusion coefficient (ADC) as well as other parameters, e.g. perfusion, as potentially early indicators of therapeutic response. Additionally in the A1 amendment this Core demonstrated the ability to map histology back to in vivo MRI voxels for animals (where whole head ex vivo MRI acquisitions are possible) using only intrinsic scan information (i.e. no implanted fiducials of any kind).