增强非线性深度学习模型通过调整混变量来估计儿科手部X射线中的骨年龄
Ki Duk Kim1, Sunggu Kyung2, Miso Jang1
1Department of Convergence Medicine, University of Ulsan College of Medicine, Asan Medical Center, 88 Olympic-Ro 43-Gil, Songpa-Gu, Seoul, 05505, Republic of Korea.
Journal of digital imaging
|June 2, 2023
概括
像性别这样的混变量改善了从X射线进行儿科骨年龄估计的深度学习模型. 通过输入或预测调整这些变量会在数据集中对模型性能产生不同的影响.
科学领域:
- 医疗成像医学成像
- 人工智能的人工智能
- 儿科放射学 儿科放射学
背景情况:
- 混变量在通用线性模型中得到调整,但在医疗任务的非线性深度学习中未得到充分探索.
- 性别是骨龄估计的关键因素,深度学习模型实现了专家级别的性能.
- 需要将混变量调整整合到非线性深度学习中,以提高骨年龄的估计.
研究的目的:
- 调查调整混变量 (性别) 在儿科骨年龄估计的非线性深度学习模型中的影响.
- 在各种深度学习架构中比较不同的混变量调整方法 (输入,输出预测,没有).
- 评估模型大小,辅助任务层次和多任务对性能和通用性的影响.
主要方法:
- 在RSNA儿科骨年龄挑战数据集上训练了基于U-Net的自动编码器,U-Net多任务学习 (MTL) 和辅助加速的MTL (AA-MTL) 模型.
- 在RSNA测试数据集上内部验证模型,并在Asan医疗中心 (AMC) 的227张儿科手部X射线上外部验证模型.
- 对模型大小,辅助任务层次和多重任务进行了废除研究,比较了混变量的调整方法.
主要成果:
- 在RSNA测试数据集中,根据输入调整混变量,在所有模型中产生了最佳性能 (MAE 5.4345.740个月).
- 在AMC外部数据集上,AA-MTL模型根据预测进行调整表现最好 (MAE 8.190个月),而其他模型则根据输入进行调整表现最好.
- 利用混变量对表现至关重要,最佳调整方法根据临床环境和模型配置而异.
结论:
- 调整混变量的方法显著影响儿童骨年龄估计的深度学习模型性能.
- 模型大小,任务层次和混调整策略对于实现最佳性能和通用性至关重要.
- 对适当的混变量调整方法的进一步研究是必要的,以开发基于深度学习的改进的骨年龄估计模型.
更多相关视频
07:29Author Spotlight: Advanced Techniques for Characterizing Tissue Mineralization in Bone Regeneration Research
Published on: September 27, 2024
809
07:12Semiautomated Longitudinal Microcomputed Tomography-based Quantitative Structural Analysis of a Nude Rat Osteoporosis-related Vertebral Fracture Model
Published on: September 28, 2017
8.2K
相关概念视频
Changes in the Appendicular Skeleton with Age
2.1K
The upper and lower limb initially develops as a small bulge called a limb bud, which appears on the lateral side of the early embryo. The upper limb bud appears near the end of the fourth week of development, with the lower limb bud appearing shortly after.
Initially, the limb buds consist of a core of mesenchyme covered by a layer of ectoderm. The ectoderm at the end of the limb bud thickens to form a narrow crest called the apical ectodermal ridge. This ridge stimulates the underlying...
Initially, the limb buds consist of a core of mesenchyme covered by a layer of ectoderm. The ectoderm at the end of the limb bud thickens to form a narrow crest called the apical ectodermal ridge. This ridge stimulates the underlying...
2.1K
Bone Remodeling
38.4K
Bone remodeling is a continuous and balanced process of bone resorption by osteoclasts and bone formation by osteoblasts. In adults, it helps maintain bone mass and calcium homeostasis. While mechanical stress can stimulate turnover as part of the normal maintenance and reparative process, several hormones also regulate bone remodeling.
38.4K
Classification of Bones
5.8K
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...
5.8K
