Collaborator: Charles Casper, PhD
Associate Professor, Department of Pediatrics
We propose to assess and catalog the quality of Magnetic Resonance Image (MRI) sessions acquired for clinical purposes and translate this assessment into an image classification method. The images were obtained from patients with early-onset (prior to age 18) demyelinating diseases seen at institutions in the Network of Pediatric Multiple Sclerosis Centers (NPMSC). Patient MRIs were transferred to the NPMSC image repository under institutional review board approval for future research and are linked with extensive clinical data characterizing disease course since onset. Due to the unstandardized acquisition of these images, their quality and usability in future research studies is currently unknown. Given the span of time and disease course that the images cover, combined with the low incidence of early-onset demyelinating diseases, it would take over ten years to build the image set prospectively and therefore the effort to store, catalog and utilize images of this nature for future research is warranted. In order to build on the image repository and make it useful to NPMSC investigators, a level of assurance that the images in the repository are of sufficient quality and type to support future studies is necessary. Implementing a classification method based on image quality measures would grow the image repository in an organized fashion so that images can be searched for usability in future studies.
BIDAC Contact: Clement Vachet
We are developing an automated framework that defines Signal and Contrast to Noise Ratio measurements (SNR & CNR) to assess quality of brain MRI images. This approach initially entails atlas-based registration and segmentation, prior to quality control measurements.
In addition, we are integrating such solution with XNAT via the python library pyXNAT. This enables us to pull data from XNAT, perform some image processing and finally upload output images or results back to XNAT.