Desiccation diagnosis in lumbar discs from clinical MRI with a probabilistic model

Abstract

Lumbar intervertebral disc diseases are among the main causes of lower back pain (LBP). Desiccation is a common disease resulting from various reasons and ultimately most people are affected by desiccation at some age. We propose a probabilistic model that incorporates intervertebral disc appearance and contextual information for automating the diagnosis of lumbar disc desiccation. We utilize a Gibbs distribution for processing localized lumbar intervertebral discs' appearance and contextual information. We use 55 clinical T2-weighted MRI for lumbar area and achieve over 96% accuracy on a cross validation experiment.