A State-of-the-Art Survey of Deep Learning for Lumbar Spine Image Analysis: X-Ray, CT, and MRI

Abstract

Article A State-of-the-Art Survey of Deep Learning for Lumbar Spine Image Analysis: X-Ray, CT, and MRI Ruyi Zhang 1,2,* 1 College of Medicine and Biological Information Engineering, Northeastern University, Chuangxin Road, Shenyang, 110016, Liaoning, China; 2390160@stu.neu.edu.cn 2 Research Institute for Medical and Biological Engineering, Ningbo University, Fenghua Road, Ningbo, 315211, Zhejiang, China Received: 17 April 2024; Revised: 12 June 2024; Accepted: 22 June 2024; Published: 17 July 2024   Abstract: Lumbar spine diseases not only endanger patients' physical health but also bring about severe psychological impacts and generate substantial medical costs. Reliable lumbar spine image analysis is crucial for diagnosing and treating lumbar spine diseases. In recent years, deep learning has rapidly developed in computer vision and medical imaging, with an increasing number of researchers applying it to the field of lumbar spine imaging. This paper studies the current state of research in deep learning applications across various modalities of lumbar spine image analysis, including X-ray, CT, and MRI. We first review the public datasets available for various tasks involving lumbar spine images. Secondly, we study the different models used in various lumbar spine image modalities (X-ray, CT, and MRI) and their applications in different tasks (classification, detection, segmentation, and reconstruction). Finally, we discuss the challenges of using deep learning in lumbar spine image analysis and provide an outlook on research and development prospects.