在医学图像分析模型中,偏见在哪里,为什么,以及如何被学习? 使用合成数据在卷积网络中的偏差编码的研究
Emma A M Stanley1, Raissa Souza1, Matthias Wilms2
1Biomedical Engineering Graduate Program, University of Calgary, Calgary, Canada; Department of Radiology, University of Calgary, Calgary, Canada; Hotchkiss Brain Institute, University of Calgary, Calgary, Canada; Alberta Children's Hospital Research Institute, University of Calgary, Calgary, Canada.
EBioMedicine
|December 13, 2024
概括
医学成像中的算法偏差深度学习是复杂的. 这项研究揭示了偏差是如何编码在模型中,发现基于强度的偏差在捷径学习中比基于形态的更有影响力.
科学领域:
- 人工智能在医学中的应用
- 医学成像分析 医学成像分析
- 深度学习 (Deep Learning) 是一种深度学习.
背景情况:
- 在医疗图像上训练的深度学习模型中的算法偏差,由于复杂的,未知的偏差来源,存在重大挑战.
- 了解这些偏见是如何在模型中编码的,对于开发可靠的医疗保健人工智能至关重要.
研究的目的:
- 研究用于医学成像的深度学习模型中的算法偏差机制.
- 为了确定医疗图像中的偏差在哪里,为什么以及如何被编码到这些模型中.
主要方法:
- 在卷积神经网络中进行系统的层wise偏差编码分析.
- 利用合成大脑磁共振成像数据,控制疾病和偏差效应.
- 在模型特征中的疾病信息和基于形态学/基于强度的偏差的量化表示.
主要成果:
- 偏见在深度学习模型中被编码,但更强大的编码不能保证捷径学习.
- 基于强度的偏见显示,当多个偏见存在时,与基于形态的偏见相比,对捷径学习的影响更大.
结论:
- 这项研究提供了对医疗成像深度学习中的算法偏差机制的基础见解.
- 控制合成数据场景对于客观地研究快捷方式学习和偏差编码是有效的.
相关概念视频
01:22Bias
3.7K
Bias refers to any tendency that prevents a question from being considered unprejudiced. In research, bias occurs when one outcome or answer is selected or encouraged over others in sampling or testing. Bias can occur during any research phase, including study design, data collection, analysis, and publication.
In statistics, a sampling bias is created when a sample is collected from a population, and some members of the population are not as likely to be chosen as others (remember, each member...
In statistics, a sampling bias is created when a sample is collected from a population, and some members of the population are not as likely to be chosen as others (remember, each member...
3.7K
01:29Bias in Epidemiological Studies
155
Biases can arise at various stages of research, from study design and data collection to analysis and interpretation. Recognizing and addressing these biases is essential to ensure the validity and reliability of epidemiological findings.Broadly speaking, biases in epidemiology fall into three main categories: selection bias, information bias, and confounding. A more detailed description of possible biases is:
155