在基于人工智能的质母细胞瘤响应评估算法中评估社会人口统计学偏差
Rachel S Lee1, Dominic LaBella2, Jikai Zhang2
1From the Duke University School of Medicine (R.S.L.); Department of Radiation Oncology (D.L.B.), Department of Radiology (K.M., E.C.), Duke University Medical Center; Department of Electrical and Computer Engineering (J.Z.), Duke University. Rachel.lee@duke.edu.
AJNR. American journal of neuroradiology
|February 9, 2026
概括
用于质母细胞瘤细分的AI模型显示出较低的人口偏差. 在多样化的数据集上训练的模型,如BraTS模型,表现更好,比在均数据上训练的模型更少的偏差.
科学领域:
- 医疗成像中的人工智能
- 神经瘤学神经瘤学
- 放射学和定量成像技术
背景情况:
- 医疗成像中的人工智能 (AI) 模型可能会表现出偏差,但根本原因尚未完全理解.
- 这项研究调查了基于AI的质母细胞瘤MRI细分中的社会人口学偏见.
- 四个nnUNet模型在不同大小和人口组成的数据集上进行了训练.
研究的目的:
- 评估AI质母细胞瘤MRI细分模型中的潜在社会人口统计偏差.
- 评估培训数据集大小和人口组成对AI模型性能的影响.
- 确定影响人工智能驱动瘤细分偏差的因素.
主要方法:
- 评估了四种人工智能模型 (FeTS2,BraTS 2024,小型同质,小型异质).
- 来自单一学术中心的480名患者的独立数据集被用于偏见评估.
- 自动细分 (FLAIR,增强瘤) 使用子得分得分;β回归分析了社会人口学影响.
主要成果:
- 在同质数据集 (白人,非西班牙裔男性) 上训练的模型显示了最低的Dice分数和显著的年龄/吸烟状态偏差.
- 这款BratS 2024型号以最小的偏差实现了最高的子得分 (0.996 FLAIR,0.999 Enhancement).
- 培训数据中的人口异质性与偏差减少相关,不论数据集大小如何.
结论:
- 质母细胞瘤MRI细分中的社会人口统计偏差AI通常很低.
- 在较小,均的数据集上训练的AI模型表现出更大的偏差.
- 在培训数据中增加的人口异质性,即使没有更大的数据集,也会减少人工智能偏差,如BraTS模型的卓越性能所示.
相关概念视频
Confirmation Biases
8.3K
The confirmation bias is the tendency to focus on information that confirms our existing beliefs and ignore information that is inconsistent with our expectations. For example, if you think that your professor is not very nice, you notice all of the instances of rude behavior exhibited by the professor while ignoring the countless pleasant interactions he is involved in on a daily basis. Have you ever fallen prey to the confirmation bias, either as the source or target of such bias?
8.3K
Hindsight Biases
4.3K
Hindsight bias leads you to believe that the event you just experienced was predictable, even though it really wasn’t. In other words, you knew all along that things would turn out the way they did. Can you relate this to the phrase "Hindsight is 20/20" now?
4.3K
Bias
7.4K
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...
7.4K
Intelligence
8.7K
The term "intelligence" is complex because it refers to both behavior and individuals, and its interpretation varies across cultures. European Americans tend to link intelligence with reasoning and cognitive skills, while in Kenya, it is tied to responsible participation in family and social life. In Uganda, intelligence is seen as the ability to know the right actions and carry them out effectively, while the Iatmul people of Papua New Guinea associate it with the capacity to remember...
8.7K
Trial and Error and Algorithm
429
A problem-solving strategy is a plan of action used to find a solution. Different strategies have distinct action plans. Trial and error involves trying different solutions until one works. For instance, to fix a broken printer, you might check ink levels, ensure the paper tray isn't jammed, and verify the printer's connection to your laptop. This method can be time-consuming but is commonly used. Thomas Edison, for example, used trial and error to find a suitable filament for the light...
429
Correspondence Bias
230
Correspondence bias, also referred to as the fundamental attribution error, describes the tendency to attribute another person’s behavior to internal characteristics rather than situational influences. This cognitive bias leads individuals to overlook external factors that may be influencing actions, thereby fostering potentially inaccurate assessments of others’ intentions and dispositions.Empirical Evidence for Correspondence BiasResearch has consistently demonstrated the...
230


