Deep Convolutional Neural Networks Using U-Net for Automatic Intervertebral Disc Segmentation in Axial MRI

Lumbar disc herniation, which occurs as a result of the rupture of the protective outer part of the disc between the vertebrae in the lumbar region and displacement of the disc due to various reasons such as a sedentary life or lifting a heavy load, compression of the disc and nerves is becoming increasingly common today and even makes life unbearable. Magnetic Resonance Imaging (MRI) technique is commonly used to diagnose lumbar disc herniation. The increase in MR images, which need to be evaluated accurately and quickly due to the excessive workload and fatigue of radiologists, also causes an increase in human error rates. To reduce human errors and assist radiologists, this study proposes a deep learning-based architecture for fast and reliable segmentation of intervertebral discs with high accuracy on MRI T2-weighted axial images. The success of segmented images is evaluated using pixel accuracy and intersection over union performance metrics, with 0.99 and 0.92 successes, respectively.