Design ensemble deep learning model for pneumonia disease classification.

IF 3.6 3区计算机科学 Q2 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

International Journal of Multimedia Information Retrieval Pub Date : 2021-01-01 Epub Date: 2021-02-20 DOI:10.1007/s13735-021-00204-7

Khalid El Asnaoui

引用次数: 0

Abstract

With the recent spread of the SARS-CoV-2 virus, computer-aided diagnosis (CAD) has received more attention. The most important CAD application is to detect and classify pneumonia diseases using X-ray images, especially, in a critical period as pandemic of covid-19 that is kind of pneumonia. In this work, we aim to evaluate the performance of single and ensemble learning models for the pneumonia disease classification. The ensembles used are mainly based on fined-tuned versions of (InceptionResNet_V2, ResNet50 and MobileNet_V2). We collected a new dataset containing 6087 chest X-ray images in which we conduct comprehensive experiments. As a result, for a single model, we found out that InceptionResNet_V2 gives 93.52% of F1 score. In addition, ensemble of 3 models (ResNet50 with MobileNet_V2 with InceptionResNet_V2) shows more accurate than other ensembles constructed (94.84% of F1 score).

Abstract Image

查看原文本刊更多论文

为肺炎疾病分类设计集合深度学习模型

随着最近 SARS-CoV-2 病毒的传播，计算机辅助诊断（CAD）受到了更多的关注。计算机辅助诊断（CAD）最重要的应用是利用 X 射线图像对肺炎疾病进行检测和分类，尤其是在属于肺炎的科维-19 病毒大流行的关键时期。在这项工作中，我们旨在评估单一学习模型和集合学习模型在肺炎疾病分类方面的性能。所使用的集合主要基于经过微调的版本（InceptionResNet_V2、ResNet50 和 MobileNet_V2）。我们收集了一个包含 6087 幅胸部 X 光图像的新数据集，并在其中进行了综合实验。结果发现，就单个模型而言，InceptionResNet_V2 的 F1 得分为 93.52%。此外，3 个模型（ResNet50、MobileNet_V2 和 InceptionResNet_V2）的集合比其他构建的集合更精确（F1 分数为 94.84%）。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

International Journal of Multimedia Information Retrieval Multiple-

CiteScore

7.80

自引率

5.40%

发文量

期刊介绍： Aims and Scope The International Journal of Multimedia Information Retrieval (IJMIR) is a scholarly archival journal publishing original, peer-reviewed research contributions. Its editorial board strives to present the most important research results in areas within the field of multimedia information retrieval. Core areas include exploration, search, and mining in general collections of multimedia consisting of information from the WWW to scientific imaging to personal archives. Comprehensive review and survey papers that offer up new insights, and lay the foundations for further exploratory and experimental work, are also relevant. Relevant topics include Image and video retrieval - theory, algorithms, and systems Social media interaction and retrieval - collaborative filtering, social voting and ranking Music and audio retrieval - theory, algorithms, and systems Scientific and Bio-imaging - MRI, X-ray, ultrasound imaging analysis and retrieval Semantic learning - visual concept detection, object recognition, and tag learning Exploration of media archives - browsing, experiential computing Interfaces - multimedia exploration, visualization, query and retrieval Multimedia mining - life logs, WWW media mining, pervasive media analysis Interactive search - interactive learning and relevance feedback in multimedia retrieval Distributed and high performance media search - efficient and very large scale search Applications - preserving cultural heritage, 3D graphics models, etc. Editorial Policies: We aim for a fast decision time (less than 4 months for the initial decision) There are no page charges in IJMIR. Papers are published on line in advance of print publication. Academic, industrial researchers, and practitioners involved with multimedia search, exploration, and mining will find IJMIR to be an essential source for important results in the field.