一个自动化的深层细分和分类模型用于腰椎椎间盘退化,并澄清其对临床决策的影响
Zafer Soydan1, Emru Bayramoglu2, Recep Karasu2
1Orthopedics and Traumatology, Bhtclinic İstanbul Tema Hospital, Nisantası University, Istanbul, Turkey.
Global spine journal
|September 12, 2023
概括
一个新的卷积神经网络 (CNN) 准确地对椎间盘退化 (IVDD) 进行分级,并改善医生的决策. 这种人工智能工具显示出高准确度和可靠性,仅仅超越了人类的评分.
科学领域:
- 医疗成像中的人工智能
- 脊柱成像分析 脊柱成像分析
- 机器学习用于疾病分类的机器学习
背景情况:
- 椎间盘退化 (IVDD) 评估对于临床决策至关重要.
- 当前的分级方法可能是主观的,耗时的.
- 开发用于IVDD分级的自动化工具是积极研究的领域.
研究的目的:
- 开发可靠的卷积神经网络 (CNN) 用于细分和分类椎间盘退化 (IVDD).
- 评估这个CNN工具对医生临床决策的影响.
- 建立一个可复制和准确的AI模型用于IVDD分级.
主要方法:
- 一个横截面数据库研究,涉及来自1137名患者的5685个磁盘.
- 使用U-net进行细分和Inception V3用于IVDD的分类.
- 在1815个磁盘上训练和测试模型,90%用于训练,10%用于测试.
- 评估模型的性能,可靠性 (ICC,kappa) 和对医生决策的影响.
主要成果:
- 实现了高细分精度 (.9597) 和分类精度 (.9346).
- 在CNN和地面真相之间表现出强烈的协议 (ICC .95-.97).
- 在CNN的帮助下,医生的成功率提高了7.9%至22%.
结论:
- 开发的自动化CNN模型在IVDD的准确性和可靠性方面明显超过了人类的分级.
- 美国有线电视新闻网的表现与现有的文献基准相提并论.
- 人工智能协助对医生的临床决策产生了积极影响,提高了诊断能力.
相关概念视频
01:27Degenerative Disc Disease I: Introduction
Degenerative disc disease is a chronic condition in which intervertebral discs gradually lose structure and function. It is not infectious or autoimmune; rather, it results from age-related biochemical and mechanical changes, influenced by genetic, metabolic, and environmental factors.Structure and Function of DiscsThe spine contains 23 intervertebral discs that absorb load, distribute forces, maintain spacing, and allow flexibility. Each disc consists of a nucleus pulposus, a gel-like core...
01:23Degenerative Disc Disease ll: Pathophysiology
The symptoms of degenerative disc disease arise from a combination of mechanical compression, vascular compromise, and biochemical inflammation, which together disrupt nerve function and produce pain.Mechanical CompressionDisc degeneration reduces height and elasticity, predisposing to herniation of the nucleus pulposus, a major cause of radicular pain. Herniations may be protrusion (bulging with intact annulus), extrusion (nucleus extends beyond disc but remains connected), or sequestration...