医学成像中的显著性驱动可解释深度学习:连接视觉可解释性与统计定量分析。

IF 4 3区 生物学 Q1 MATHEMATICAL & COMPUTATIONAL BIOLOGY Biodata Mining Pub Date : 2024-06-22 DOI:10.1186/s13040-024-00370-4
Yusuf Brima, Marcellin Atemkeng
{"title":"医学成像中的显著性驱动可解释深度学习:连接视觉可解释性与统计定量分析。","authors":"Yusuf Brima, Marcellin Atemkeng","doi":"10.1186/s13040-024-00370-4","DOIUrl":null,"url":null,"abstract":"<p><p>Deep learning shows great promise for medical image analysis but often lacks explainability, hindering its adoption in healthcare. Attribution techniques that explain model reasoning can potentially increase trust in deep learning among clinical stakeholders. In the literature, much of the research on attribution in medical imaging focuses on visual inspection rather than statistical quantitative analysis.In this paper, we proposed an image-based saliency framework to enhance the explainability of deep learning models in medical image analysis. We use adaptive path-based gradient integration, gradient-free techniques, and class activation mapping along with its derivatives to attribute predictions from brain tumor MRI and COVID-19 chest X-ray datasets made by recent deep convolutional neural network models.The proposed framework integrates qualitative and statistical quantitative assessments, employing Accuracy Information Curves (AICs) and Softmax Information Curves (SICs) to measure the effectiveness of saliency methods in retaining critical image information and their correlation with model predictions. Visual inspections indicate that methods such as ScoreCAM, XRAI, GradCAM, and GradCAM++ consistently produce focused and clinically interpretable attribution maps. These methods highlighted possible biomarkers, exposed model biases, and offered insights into the links between input features and predictions, demonstrating their ability to elucidate model reasoning on these datasets. Empirical evaluations reveal that ScoreCAM and XRAI are particularly effective in retaining relevant image regions, as reflected in their higher AUC values. However, SICs highlight variability, with instances of random saliency masks outperforming established methods, emphasizing the need for combining visual and empirical metrics for a comprehensive evaluation.The results underscore the importance of selecting appropriate saliency methods for specific medical imaging tasks and suggest that combining qualitative and quantitative approaches can enhance the transparency, trustworthiness, and clinical adoption of deep learning models in healthcare. This study advances model explainability to increase trust in deep learning among healthcare stakeholders by revealing the rationale behind predictions. Future research should refine empirical metrics for stability and reliability, include more diverse imaging modalities, and focus on improving model explainability to support clinical decision-making.</p>","PeriodicalId":48947,"journal":{"name":"Biodata Mining","volume":null,"pages":null},"PeriodicalIF":4.0000,"publicationDate":"2024-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11193223/pdf/","citationCount":"0","resultStr":"{\"title\":\"Saliency-driven explainable deep learning in medical imaging: bridging visual explainability and statistical quantitative analysis.\",\"authors\":\"Yusuf Brima, Marcellin Atemkeng\",\"doi\":\"10.1186/s13040-024-00370-4\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Deep learning shows great promise for medical image analysis but often lacks explainability, hindering its adoption in healthcare. Attribution techniques that explain model reasoning can potentially increase trust in deep learning among clinical stakeholders. In the literature, much of the research on attribution in medical imaging focuses on visual inspection rather than statistical quantitative analysis.In this paper, we proposed an image-based saliency framework to enhance the explainability of deep learning models in medical image analysis. We use adaptive path-based gradient integration, gradient-free techniques, and class activation mapping along with its derivatives to attribute predictions from brain tumor MRI and COVID-19 chest X-ray datasets made by recent deep convolutional neural network models.The proposed framework integrates qualitative and statistical quantitative assessments, employing Accuracy Information Curves (AICs) and Softmax Information Curves (SICs) to measure the effectiveness of saliency methods in retaining critical image information and their correlation with model predictions. Visual inspections indicate that methods such as ScoreCAM, XRAI, GradCAM, and GradCAM++ consistently produce focused and clinically interpretable attribution maps. These methods highlighted possible biomarkers, exposed model biases, and offered insights into the links between input features and predictions, demonstrating their ability to elucidate model reasoning on these datasets. Empirical evaluations reveal that ScoreCAM and XRAI are particularly effective in retaining relevant image regions, as reflected in their higher AUC values. However, SICs highlight variability, with instances of random saliency masks outperforming established methods, emphasizing the need for combining visual and empirical metrics for a comprehensive evaluation.The results underscore the importance of selecting appropriate saliency methods for specific medical imaging tasks and suggest that combining qualitative and quantitative approaches can enhance the transparency, trustworthiness, and clinical adoption of deep learning models in healthcare. This study advances model explainability to increase trust in deep learning among healthcare stakeholders by revealing the rationale behind predictions. Future research should refine empirical metrics for stability and reliability, include more diverse imaging modalities, and focus on improving model explainability to support clinical decision-making.</p>\",\"PeriodicalId\":48947,\"journal\":{\"name\":\"Biodata Mining\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":4.0000,\"publicationDate\":\"2024-06-22\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11193223/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Biodata Mining\",\"FirstCategoryId\":\"99\",\"ListUrlMain\":\"https://doi.org/10.1186/s13040-024-00370-4\",\"RegionNum\":3,\"RegionCategory\":\"生物学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"MATHEMATICAL & COMPUTATIONAL BIOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Biodata Mining","FirstCategoryId":"99","ListUrlMain":"https://doi.org/10.1186/s13040-024-00370-4","RegionNum":3,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"MATHEMATICAL & COMPUTATIONAL BIOLOGY","Score":null,"Total":0}
引用次数: 0

摘要

深度学习在医学图像分析方面大有可为,但往往缺乏可解释性,阻碍了其在医疗保健领域的应用。解释模型推理的归因技术有可能增加临床利益相关者对深度学习的信任。本文提出了一个基于图像的显著性框架,以增强深度学习模型在医学图像分析中的可解释性。我们使用基于路径的自适应梯度积分、无梯度技术和类激活映射及其衍生物,对最近的深度卷积神经网络模型从脑肿瘤 MRI 和 COVID-19 胸部 X 光数据集中得出的预测结果进行归因。所提出的框架综合了定性和统计定量评估,使用准确度信息曲线(AIC)和软最大信息曲线(SIC)来衡量突出度方法在保留关键图像信息方面的有效性及其与模型预测的相关性。目测结果表明,ScoreCAM、XRAI、GradCAM 和 GradCAM++ 等方法能持续生成重点突出、临床可解释的归因图。这些方法突出了可能的生物标记物,暴露了模型偏差,并提供了输入特征与预测之间联系的见解,证明了它们在这些数据集上阐明模型推理的能力。经验评估显示,ScoreCAM 和 XRAI 在保留相关图像区域方面特别有效,这反映在它们较高的 AUC 值上。结果强调了为特定医学成像任务选择合适的突出度方法的重要性,并表明结合定性和定量方法可以提高深度学习模型在医疗保健领域的透明度、可信度和临床应用。本研究通过揭示预测背后的原理,提高了模型的可解释性,从而增加了医疗保健利益相关者对深度学习的信任。未来的研究应完善稳定性和可靠性的经验指标,纳入更多不同的成像模式,并侧重于提高模型的可解释性,以支持临床决策。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
Saliency-driven explainable deep learning in medical imaging: bridging visual explainability and statistical quantitative analysis.

Deep learning shows great promise for medical image analysis but often lacks explainability, hindering its adoption in healthcare. Attribution techniques that explain model reasoning can potentially increase trust in deep learning among clinical stakeholders. In the literature, much of the research on attribution in medical imaging focuses on visual inspection rather than statistical quantitative analysis.In this paper, we proposed an image-based saliency framework to enhance the explainability of deep learning models in medical image analysis. We use adaptive path-based gradient integration, gradient-free techniques, and class activation mapping along with its derivatives to attribute predictions from brain tumor MRI and COVID-19 chest X-ray datasets made by recent deep convolutional neural network models.The proposed framework integrates qualitative and statistical quantitative assessments, employing Accuracy Information Curves (AICs) and Softmax Information Curves (SICs) to measure the effectiveness of saliency methods in retaining critical image information and their correlation with model predictions. Visual inspections indicate that methods such as ScoreCAM, XRAI, GradCAM, and GradCAM++ consistently produce focused and clinically interpretable attribution maps. These methods highlighted possible biomarkers, exposed model biases, and offered insights into the links between input features and predictions, demonstrating their ability to elucidate model reasoning on these datasets. Empirical evaluations reveal that ScoreCAM and XRAI are particularly effective in retaining relevant image regions, as reflected in their higher AUC values. However, SICs highlight variability, with instances of random saliency masks outperforming established methods, emphasizing the need for combining visual and empirical metrics for a comprehensive evaluation.The results underscore the importance of selecting appropriate saliency methods for specific medical imaging tasks and suggest that combining qualitative and quantitative approaches can enhance the transparency, trustworthiness, and clinical adoption of deep learning models in healthcare. This study advances model explainability to increase trust in deep learning among healthcare stakeholders by revealing the rationale behind predictions. Future research should refine empirical metrics for stability and reliability, include more diverse imaging modalities, and focus on improving model explainability to support clinical decision-making.

求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
Biodata Mining
Biodata Mining MATHEMATICAL & COMPUTATIONAL BIOLOGY-
CiteScore
7.90
自引率
0.00%
发文量
28
审稿时长
23 weeks
期刊介绍: BioData Mining is an open access, open peer-reviewed journal encompassing research on all aspects of data mining applied to high-dimensional biological and biomedical data, focusing on computational aspects of knowledge discovery from large-scale genetic, transcriptomic, genomic, proteomic, and metabolomic data. Topical areas include, but are not limited to: -Development, evaluation, and application of novel data mining and machine learning algorithms. -Adaptation, evaluation, and application of traditional data mining and machine learning algorithms. -Open-source software for the application of data mining and machine learning algorithms. -Design, development and integration of databases, software and web services for the storage, management, retrieval, and analysis of data from large scale studies. -Pre-processing, post-processing, modeling, and interpretation of data mining and machine learning results for biological interpretation and knowledge discovery.
期刊最新文献
Deep joint learning diagnosis of Alzheimer's disease based on multimodal feature fusion. Modeling heterogeneity of Sudanese hospital stay in neonatal and maternal unit: non-parametric random effect models with Gamma distribution. Ensemble feature selection and tabular data augmentation with generative adversarial networks to enhance cutaneous melanoma identification and interpretability. Priority-Elastic net for binary disease outcome prediction based on multi-omics data. A regularized Cox hierarchical model for incorporating annotation information in predictive omic studies.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1