主体级脊柱骨质疏松性骨折预测,结合深度学习的脊椎输出和有限的人口数据
Nathan M Cross1, Jessica Perry2, Qifei Dong3
1Department of Radiology, University of Washington, 1959 NE Pacific Street, Box 357115, Seattle, WA, 98195-7115, USA. nmcross@uw.edu.
Archives of osteoporosis
|September 10, 2024
概括
使用深度学习进行的脊椎骨折自动查实现了高精度 (AUC-ROC 0.968). 最好的模型结合了最大的脊椎骨折得分和年龄,以可靠的患者水平骨折预测.
科学领域:
- 放射学 放射学是一门学科.
- 人工智能的人工智能
- 骨质疏松症研究 骨质疏松症研究
背景情况:
- 骨质疏松性脊椎骨折很普遍,并导致显著的发病率.
- 通过X光扫描对偶然的脊椎骨折进行自动查,可以提高骨质疏松症的检测和管理.
- 放射图片代表了体积最大的医学成像模式.
研究的目的:
- 开发和比较患者层面的骨折预测模型,使用逻辑回归 (LR) 和通用添加模型 (GAM).
- 评估包含深度学习衍生的脊椎骨折得分和人口统计共变量的模型的性能.
- 通过准确和总结的患者水平骨折预测来指导临床管理.
主要方法:
- 通过使用脊椎骨折预测和人口统计数据 (年龄,种族/族裔) 以后期创建主体级LR和GAM模型.
- 从669个受试者的1176张X射线图中对8663个脊椎的分类器输出进行了分层的五倍交叉验证.
- 使用接收器操作特征 (ROC) 和精度回忆 (PR) 曲线进行模型评估.
主要成果:
- 性能最好的模型是GAM,它包含了三个最高的最大脊椎骨折得分和年龄,达到0.968.96的AUC-ROC.
- 使用排名最高的脊椎分数提高了模型性能,而不是使用LR和GAM的所有分数.
- 包括年龄在内略有提高了GAM表现,而种族并没有显著影响结果.
结论:
- 最大脊椎骨折分数是自动骨折查的高性能模型中的关键预测指标.
- 主体级模型有效地预测骨折,即使结合多个脊椎预测,保持良好的预测性能.
- 可调节的值策略允许临床控制骨折检测中的灵敏度和特异性.
相关概念视频
01:18Classification of Bones
5.2K
The bones of the human skeletal system are of varied shapes, sizes, and functions. They can be classified based on their shape and function into four major classes: long bones, short bones, flat bones, and irregular bones. Some classifications include a fifth type, the sesamoid bones, as a separate class, whereas others categorize them under short bones.
Long and Short Bones
The appendicular skeleton, particularly the upper and lower limbs, is primarily made of long and short bones. The...
Long and Short Bones
The appendicular skeleton, particularly the upper and lower limbs, is primarily made of long and short bones. The...
5.2K
01:16Vertebral Column: Regions and Curvature
3.5K
The vertebral column or spine is a flexible column that supports the head, neck, and body and allows for their movements. It also protects the spinal cord.
Regions of the Vertebral Column
In an adult, the spine is subdivided into five regions: the cervical, the thoracic, the lumbar, the sacral, and the coccygeal region. The spine initially develops as a series of 33 vertebrae; after 20 years of age, the nine bones in the sacral region, five sacral, and four coccygeal bones fuse to form...
Regions of the Vertebral Column
In an adult, the spine is subdivided into five regions: the cervical, the thoracic, the lumbar, the sacral, and the coccygeal region. The spine initially develops as a series of 33 vertebrae; after 20 years of age, the nine bones in the sacral region, five sacral, and four coccygeal bones fuse to form...
3.5K