MedShift: Automated Identification of Shift Data for Medical Image Dataset Curation

MedShift: Automated Identification of Shift Data for Medical Image Dataset Curation
复制标题

MedShift:自动识别医学图像数据集管理的移位数据

DOI:
10.1109/jbhi.2023.3275104
复制
发表时间:
2023
影响因子:
7.7
通讯作者:
Banerjee, Imon
Banerjee, Imon
中科院分区:
工程技术1区
文献类型:
--
作者:
Guo, Xiaoyuan;Gichoya, Judy Wawira;Trivedi, Hari;Purkayastha, Saptarshi;Banerjee, Imon

文献摘要

参考文献

相似文献

长期以来,医疗领域对噪声外部数据的自动化管理一直有很高的需求,因为人工智能技术需要使用各种来源的干净的注释数据进行验证。识别内部和外部来源之间的差异是管理高质量数据集的基本步骤,因为来自不同来源的数据分布可能会有很大差异,并随后影响AI模型的性能。检测数据偏移的主要挑战是:(1)访问医疗机构的私人数据进行手动检测;(2)缺乏自动化方法来学习有效的偏移数据表示,而无需训练样本。为了克服这些问题,我们提出了一个自动化的管道称为MedShiftto检测顶级移位样本和评估移位数据的重要性,而无需在内部和外部组织之间共享数据。MedShiftemploys无监督的异常检测器来学习内部分布和识别外部数据集显示显著移位的样本,然后比较它们的性能。为了量化检测到的移位数据的影响,我们训练了一个多类分类器,该分类器学习内部域知识,并在丢弃移位数据后评估外部域中每个类的分类性能。我们还提出了数据质量指标来量化内部和外部数据集之间的差异。我们使用来自多个外部来源的肌肉骨骼X线片(MURA)和胸部X线数据集验证MedShift的有效性。我们的实验表明,我们提出的移位数据检测管道可以帮助医疗中心更有效地管理高质量的数据集。
Automated curation of noisy external data in the medical domain has long been in high demand, as AI technologies need to be validated using various sources with clean, annotated data. Identifying the variance between internal and external sources is a fundamental step in curating a high-quality dataset, as the data distributions from different sources can vary significantly and subsequently affect the performance of AI models. The primary challenges for detectingdata shiftsare - (1) accessing private data across healthcare institutions for manual detection and (2) the lack of automated approaches to learn efficient shift-data representation without training samples. To overcome these problems, we propose an automated pipeline calledMedShiftto detect top-level shift samples and evaluate the significance of shift data without sharing data between internal and external organizations.MedShiftemploys unsupervised anomaly detectors to learn the internal distribution and identify samples showing significant shiftness for external datasets, and then compares their performance. To quantify the effects of detected shift data, we train a multi-class classifier that learns internal domain knowledge and evaluates the classification performance for each class in external domains after dropping the shift data. We also propose adata quality metricto quantify the dissimilarity between internal and external datasets. We verify the efficacy ofMedShiftusing musculoskeletal radiographs (MURA) and chest X-ray datasets from multiple external sources. Our experiments show that our proposed shift data detection pipeline can be beneficial for medical centers to curate high-quality datasets more efficiently.
临床前和临床多模态成像研究的数据管理
DOI: 10.1007/s11307-019-01339-0
发表时间: 2019
影响因子: 3.1
作者:
G. G. Yamoah;Liji Cao;Chao Wu;F. Beekman;B. Vandeghinste;J. Mannheim;S. Rosenhain;Kevin Leonardic;F. Kiessling;F. Gremse
通讯作者: F. Gremse
DOI: 10.1016/j.jvir.2019.05.026
发表时间: 2020-01-01
影响因子: 2.9
作者:
Ni, Jason C.;Shpanskaya, Katie;Wang, David S.
通讯作者: Wang, David S.
CVAD - 无监督图像异常检测器
DOI: --
发表时间: 2021
期刊: Softw. Impacts
影响因子: --
作者:
X. Guo;J. Gichoya;S. Purkayastha;Imon Banerjee
通讯作者: Imon Banerjee
DOI: 10.1186/1471-2458-14-1144
发表时间: 2014-11-05
期刊: BMC public health
影响因子: 4.5
作者:
van Panhuis WG;Paul P;Emerson C;Grefenstette J;Wilder R;Herbst AJ;Heymann D;Burke DS
通讯作者: Burke DS