Automated quantification of intervertebral disc volume loss, ADC and normalized T2 using convolutional neural networks

Intervertebral disc degeneration is the leading cause of chronic low back pain (CLBP). However, there are no objective measures of the disc degeneration process. The current gold standard for grading disc degeneration relies on visual assessment of discs using MRI, which may not adequately capture complex physiological changes in degenerating discs. We developed a deep convolutional neural network to automatically segment vertebral structures and calculate quantitative MRI metrics. Results show significant changes in ADC and normalized T2 values with disc volume loss.