Automatic Detection of Coronavirus Disease (COVID-19) in X-ray and CT Images: A Machine Learning-Based Approach (2004.10641v1)

Abstract: The newly identified Coronavirus pneumonia, subsequently termed COVID-19, is highly transmittable and pathogenic with no clinically approved antiviral drug or vaccine available for treatment. The most common symptoms of COVID-19 are dry cough, sore throat, and fever. Symptoms can progress to a severe form of pneumonia with critical complications, including septic shock, pulmonary edema, acute respiratory distress syndrome and multi-organ failure. While medical imaging is not currently recommended in Canada for primary diagnosis of COVID-19, computer-aided diagnosis systems could assist in the early detection of COVID-19 abnormalities and help to monitor the progression of the disease, potentially reduce mortality rates. In this study, we compare popular deep learning-based feature extraction frameworks for automatic COVID-19 classification. To obtain the most accurate feature, which is an essential component of learning, MobileNet, DenseNet, Xception, ResNet, InceptionV3, InceptionResNetV2, VGGNet, NASNet were chosen amongst a pool of deep convolutional neural networks. The extracted features were then fed into several machine learning classifiers to classify subjects as either a case of COVID-19 or a control. This approach avoided task-specific data pre-processing methods to support a better generalization ability for unseen data. The performance of the proposed method was validated on a publicly available COVID-19 dataset of chest X-ray and CT images. The DenseNet121 feature extractor with Bagging tree classifier achieved the best performance with 99% classification accuracy. The second-best learner was a hybrid of the a ResNet50 feature extractor trained by LightGBM with an accuracy of 98%.

• The paper achieves up to 99% accuracy in COVID-19 detection using deep CNNs and a Bagging tree classifier.
• It employs transfer learning with multiple CNN architectures like DenseNet, MobileNet, and Xception to extract robust features.
• The study highlights the potential for rapid, automated diagnostics to reduce clinical delays during the pandemic.

Overview of Automatic Detection of COVID-19 in X-ray and CT Images Using Machine Learning

Introduction

The paper "Automatic Detection of Coronavirus Disease (COVID-19) in X-ray and CT Images: A Machine Learning Based Approach" presents a systematic paper of image-based detection methods for COVID-19 using machine learning techniques. This paper was undertaken in response to the urgent need for rapid and accurate diagnostic methods amidst the COVID-19 pandemic, where traditional diagnostic methods like RT-PCR tests faced logistical challenges and delays. The authors aim to leverage the inherent potential of medical imaging and machine learning to facilitate timely detection and monitoring.

Methodology and Experimental Setup

The authors employed a comprehensive analysis using multiple deep convolutional neural networks (CNNs) for feature extraction and a suite of machine learning classifiers for categorizing X-ray and CT images. Notably, the adopted CNN architectures include MobileNet, DenseNet, Xception, among others, primarily selected for their prominence in capturing image features effectively. DenseNet121 coupled with a Bagging tree classifier emerged as the most effective, achieving a classification accuracy of 99%. A key strategy deployed was transfer learning, utilizing pre-trained CNN models thus mitigating the risk of overfitting given the limited dataset size.

The dataset compiled consists of 117 X-ray and 20 CT images of COVID-19 positive cases complemented by an equivalent number of normal cases derived from existing datasets on Kaggle. This balanced dataset facilitated robust training and evaluation using 10-fold cross-validation.

Results and Discussions

The DenseNet121 architecture combined with Bagging tree classifiers demonstrated superior classification performance, indicating that the architecture could offer precise and reliable diagnosis from medical images. As the second-best performer, ResNet50 with LightGBM, achieved an accuracy of 98%. These results highlight that deep features extracted via CNNs significantly enhance the model's capability to correctly distinguish COVID-19 positive cases from healthy controls.

The authors reported efficient computational times for feature extraction and model training, which is critical for real-world deployment scenarios where time-constrained diagnosis is essential. The discussion acknowledges some limitations, particularly the dependency on a limited dataset, which poses challenges in generalizability. However, the results provide a promising starting point for large-scale clinical validation.

Implications and Future Scope

This research has immediate implications for the development of computer-aided diagnosis systems capable of alleviating the burden on healthcare systems by enabling automated, fast, and accurate COVID-19 screening tools. The success of using pre-trained CNNs for feature extraction underscores the value of integrating deep learning within diagnostic workflows, particularly for emerging diseases with rapidly evolving datasets.

Future research efforts would benefit from expanding the dataset with a wider range of clinical images, incorporating diverse patient demographics and varying disease severities to enhance model robustness and generalization. Adapting the model for real-time clinical environments remains a pertinent direction, necessitating further optimization of the inference pipeline to ensure compliance with clinical standards and regulatory requirements.

In conclusion, this paper lays a robust groundwork for ongoing and future efforts in employing machine learning for medical diagnosis, not only for COVID-19 but extending across various domains of infectious diseases where rapid diagnostics are critically required.