School of Medicine, Shanghai Jiao Tong University, 200025, Shanghai, PR China.
School of Traffic and Transportation Engineering, Central South University, 410075, Hunan, PR China.
Comput Biol Med. 2021 Apr;131:104252. doi: 10.1016/j.compbiomed.2021.104252. Epub 2021 Feb 2.
Chest X-ray radiography (CXR) has been widely considered as an accessible, feasible, and convenient method to evaluate suspected patients' lung involvement during the COVID-19 pandemic. However, with the escalating number of suspected cases, traditional diagnosis via CXR fails to deliver results within a short period of time. Therefore, it is crucial to employ artificial intelligence (AI) to enhance CXRs for obtaining quick and accurate diagnoses. Previous studies have reported the feasibility of utilizing deep learning methods to screen for COVID-19 using CXR and CT results. However, these models only use a single deep learning network for chest radiograph detection; the accuracy of this approach required further improvement.
In this study, we propose a three-step hybrid ensemble model, including a feature extractor, a feature selector, and a classifier. First, a pre-trained AlexNet with an improved structure extracts the original image features. Then, the ReliefF algorithm is adopted to sort the extracted features, and a trial-and-error approach is used to select the n most important features to reduce the feature dimension. Finally, an SVM classifier provides classification results based on the n selected features.
Compared to five existing models (InceptionV3: 97.916 ± 0.408%; SqueezeNet: 97.189 ± 0.526%; VGG19: 96.520 ± 1.220%; ResNet50: 97.476 ± 0.513%; ResNet101: 98.241 ± 0.209%), the proposed model demonstrated the best performance in terms of overall accuracy rate (98.642 ± 0.398%). Additionally, compared to the existing models, the proposed model demonstrates a considerable improvement in classification time efficiency (SqueezeNet: 6.602 ± 0.001s; InceptionV3: 12.376 ± 0.002s; ResNet50: 10.952 ± 0.001s; ResNet101: 18.040 ± 0.002s; VGG19: 16.632 ± 0.002s; proposed model: 5.917 ± 0.001s).
The model proposed in this article is practical and effective, and can provide high-precision COVID-19 CXR detection. We demonstrated its suitability to aid medical professionals in distinguishing normal CXRs, viral pneumonia CXRs and COVID-19 CXRs efficiently on small sample sizes.
在 COVID-19 大流行期间，胸部 X 射线摄影（CXR）被广泛认为是一种易于获得、可行且方便的方法，可用于评估疑似患者的肺部受累情况。然而，随着疑似病例数量的不断增加，传统的 CXR 诊断方法无法在短时间内得出结果。因此，利用人工智能（AI）增强 CXR 以快速准确诊断至关重要。先前的研究报告了使用 CXR 和 CT 结果利用深度学习方法筛查 COVID-19 的可行性。然而，这些模型仅使用单个深度学习网络进行胸部 X 射线检测；该方法的准确性需要进一步提高。
在这项研究中，我们提出了一种三步骤混合集成模型，包括特征提取器、特征选择器和分类器。首先，使用经过改进结构的预训练 AlexNet 提取原始图像特征。然后，采用 ReliefF 算法对提取的特征进行排序，并采用试错法选择最重要的 n 个特征来降低特征维度。最后，SVM 分类器根据 n 个选择的特征提供分类结果。
与五个现有模型（InceptionV3：97.916 ± 0.408%；SqueezeNet：97.189 ± 0.526%；VGG19：96.520 ± 1.220%；ResNet50：97.476 ± 0.513%；ResNet101：98.241 ± 0.209%）相比，所提出的模型在总准确率（98.642 ± 0.398%）方面表现最佳。此外，与现有模型相比，所提出的模型在分类时间效率方面有了相当大的提高（SqueezeNet：6.602 ± 0.001s；InceptionV3：12.376 ± 0.002s；ResNet50：10.952 ± 0.001s；ResNet101：18.040 ± 0.002s；VGG19：16.632 ± 0.002s；所提出的模型：5.917 ± 0.001s）。
本文提出的模型实用有效，可提供高精度的 COVID-19 CXR 检测。我们证明了它适用于帮助医疗专业人员在小样本量上有效区分正常 CXR、病毒性肺炎 CXR 和 COVID-19 CXR。
