Case-Control Sampling for Brain Imaging and Enhancement Prediction

• Activity Number: 493
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Gina-Maria Pomann*+ and Russell Shinohara and Ana-Maria Staicu and Elizabeth Sweeney and Daniel Reich
• Companies: North Carolina State University and Univ of Pennsylvania and North Carolina State University and Johns Hopkins University and National Institute of Neurological Disorders & Stroke
• Keywords: Magnetic Resonance Imaging ; Case-Control Sampling ; Rare Events Data ; Brain Imaging ; Logistic Regression

Abstract

Multiple sclerosis (MS) is a disease associated with inflammation of the brain and causes morbidity and disability. Irregular blood flow in white matter of the brain is associated with clinical relapses in patients with MS. This abnormality in active MS lesions is commonly evaluated through identification of enhancing voxels on Magnetic Resonance (MR) images. Current clinical practice includes intravenous injection of contrast agents that are occasionally toxic to the patient and can increase the cost of imaging by over 40 \%. Local image regression methodology has recently been developed to predict enhancing MS lesions without using such contrast agents. We extend this model to account for the rarity of enhancement as well as to incorporate historical information about lesions. A covariate describing such historical information is added to the model and found to substantially improve prediction. We consider 77 brain MR imaging studies on 15 patients which include historical imaging covariates along with T1-w, T2-w images. Additionally, we present a scan-stratified case-control sampling technique that accounts for the rarity of enhancement while also reducing computational cost.