Automatically Diagnosing Disk Bulge and Disk Herniation With Lumbar Magnetic Resonance Images by Using Deep Convolutional Neural Networks: Method Development Study (Preprint)

• DOI: 10.2196/preprints.14755
• Publication Date: 2019-05-21T18:34:45Z
• Authors: Qiong Pan; Kai Zhang; Lin He; Zhou Dong; Lei Zhang; Xiaohang Wu; Yi Wu; Yanjun Gao
• Keywords: Intervertebral disk; Vertebra; Back Pain

<sec> <title>BACKGROUND</title> Disk herniation and disk bulge are two common disorders of lumbar intervertebral disks (IVDs) that often result in numbness, pain the lower limbs, back pain. Magnetic resonance (MR) imaging is one most efficient techniques for detecting diseases widely used making clinical diagnoses at hospitals. However, there a lack tools effectively interpreting massive amounts MR images to meet requirements many radiologists. </sec> <title>OBJECTIVE</title> The aim this study was present an automatic system diagnosing saves time can significantly reduce workload <title>METHODS</title> diagnosis vertebral highly dependent on medical images. Therefore, we chose diseases—disk herniation—as research subjects. This mainly about identifying position IVDs (lumbar vertebra [L] 1 L2, L2-L3, L3-L4, L4-L5, L5 sacral [S] 1) by analyzing geometrical relationship between sagittal axial classifying via deep convolutional neural networks. <title>RESULTS</title> involved 4 steps. In first step, it automatically located bodies (including L1, L3, L4, L5, S1) using faster region-based network, our fourfold cross-validation showed 100% accuracy. second spontaneously identified corresponding each image with third accuracy locating region interest 100%. fourth 3-class classification (normal disk, bulge, herniation) accuracies L1-L2, L5-S1 were 92.7%, 84.4%, 92.1%, 90.4%, 84.2%, respectively. <title>CONCLUSIONS</title> successfully built, could classify normal disks, herniation. provided web-based test improve diagnostic efficiency standardized reports. also be detect other abnormalities cervical spondylosis.