Evaluating Convolutional Neural Networks for COVID-19 classification in chest X-ray images

Leonardo Gabriel Ferreira Rodrigues, Larissa Ferreira Rodrigues, Danilo Ferreira da Silva, João Fernando Mari · 2024

Read Evaluating Convolutional Neural Networks for COVID-19 classification in chest X-ray images by Leonardo Gabriel Ferreira Rodrigues, Larissa Ferreira Rodrigues, Danilo Ferreira da Silva, João Fernando Mari (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-19362.txt.

Introduction

Leonardo Gabriel Ferreira Rodrigues and co-authors compare neural networks for spotting COVID-19 in chest X-rays, in this 2024 paper. During the pandemic, fast screening mattered, and X-ray images were widely shared. The team tested whether standard image networks could sort positive from negative cases with little labeled data.

Four networks were trained: AlexNet, VGG-11, SqueezeNet, and DenseNet-121. Training used ten-fold cross-validation, gentle fine-tuning, and image augmentation to stretch the small set of positive scans. All four networks scored above 97 percent accuracy. SqueezeNet, the smallest of the set, did best at 99.20 percent. The authors say such tools could aid screening where expert readers are scarce, while the high scores rest on one shared image collection and need checks on wider data.

Catalog record

CREATORLeonardo Gabriel Ferreira Rodrigues, Larissa Ferreira Rodrigues, Danilo Ferreira da Silva, João Fernando Mari
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-19362

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Evaluating Convolutional Neural Networks for COVID-19 classification in chest X-ray images Leonardo Gabriel Ferreira Rodrigues, Larissa Ferreira Rodrigues, Danilo Ferreira da Silva, João Fernando Mari Abstract † † The definite version of this paper was published by SBC-OpenLib as part of the WVC 2020 proceedings. DOI: https://doi.org/10.5753/wvc.2020.13480 Coronavirus Disease 2019 (COVID-19) pandemic rapidly spread globally, impacting the lives of billions of people. The effective screening of infected patients is a critical step to struggle with COVID-19, and treating the patients avoiding this quickly disease spread. The need for automated and scalable methods has increased due to the unavailability of accurate automated toolkits. Recent researches using chest X-ray images suggest they include relevant information about the COVID-19 virus. Hence, applying machine learning techniques combined with radiological imaging promises to identify this disease accurately. It is straightforward to collect these images once it is spreadly shared and analyzed in the world. This paper presents a method for automatic COVID-19 detection using chest X-ray images through four convolutional neural networks, namely: AlexNet, VGG-11, SqueezeNet, and DenseNet-121. This method had been providing accurate diagnostics for positive or negative COVID-19 classification. We validate our experiments using a ten-fold cross-validation procedure over the training and test sets. Our findings include the shallow fine-tuning and data augmentation strategies that can assist in dealing with the low number of positive COVID-19 images publicly available. The accuracy for all CNNs is higher than 97.00%, and the SqueezeNet model achieved the best result with 99.20%. I Introduction Coronavirus Disease 2019 (COVID-19) caused by a novel coronavirus, officially named Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) [ 1 ] , is a pandemic that first emerged in the Chinese city of Wuhan, and rapidly spread to other countries, affecting Italy, Iran, Spain, Brazil, Russia, India, and United States severely. All countries affected by COVID-19 imposed a nationwide lockdown in an attempt to slow the spread of the virus, causing a profound overall impact on the lives of billions of people from a health, safety, and economic perspective [ 2 ] [ 3 ] . The COVID-19 can cause illness to the respiratory system leading to inflammation of the lungs and pneumonia [ 4 ] . There is no known specific therapeutic drugs or vaccine for COVID-19, and the impact in the healthcare system is also high due to the number of people that needs intensive care unit (ICU) admission and breathing machine for long periods [ 5 ] . The most common test technique currently used for COVID-19 diagnosis is the reverse transcription-polymerase chain reaction (RT-PCR). However, considering the difficulties of distributing the kits and collecting the samples and the waiting time for results, auxiliary diagnostics methods are welcome to assist the medical team decision making. In this context, the development of computer-aided diagnosis systems based on machine learning is essential and widely applied in several fields of medicine [ 6 ] [ 7 ] [ 8 ] . Early studies demonstrated that many patients infected with COVID-19 present abnormalities in chest X-ray images [ 9 ] [ 10 ] [ 11 ] . These images can be easily collected, shared, and analyzed around the world. Moreover, the task of COVID-19 identification is not easy, and the specialist reviewing the chest X-ray needs to look for white patches in the lungs, i.e., air sacs filled with pus or water. However, these white patches can also be confused with diseases such as tuberculosis, bronchitis, and other types of pneumonia caused by different viruses or bacteria. In this study, we aim to explore the identification of COVID-19 using chest X-ray images due to its reduced cost, fast result, and general availability. Our principal goal is to reach the best possible identification rate among COVID-19 and other types of pneumonia. We applied a pure deep learning approach comparing four Convolutional Neural Networks (CNNs): AlexNet, VGG-11, SqueezeNet and DenseNet-121, and we evaluated the performance using a k-fold cross-validation procedure over the training and test sets. Moreover, we carried a confusion matrix analysis to measure accuracy, precision, recall, and F1-score indices. Furthermore, the chest X-ray image dataset used in this work contains a few positive images on the COVID-19. To deal with this, we applied Shallow Fine-Tuning (SFT) training and data augmentation based on random rotation and shifting to balance the class distribution and the performance of the classification. In addition, our approach is fast and simple producing high performing system. As far as we know, our result is the best obtained for COVID-19 identification in chest X-ray images. We believe that our proposed method can contribute to future researches intended to help healthcare workers to identify COVID-19 and to manage patient’s conditions. The remaining of this paper is organized as follows: Section II surveys related work; Section III describes the material and methods; In Section IV we present and discuss the results obtained. Finally, conclusions and future work are presented in Section V . II Related Work The COVID-19 has been attracting much attention from the image analysis research community due to its severity. In this sense, Narin et al. [ 12 ] compared three different CNN architectures (ResNet50, Inception-V3, and InceptionResNetV2) to identify COVID-19 in chest X-ray images. They used a dataset composed of fifty COVID-19 images taken from the open-source GitHub repository shared by Dr. Joseph Cohen [ 13 ] and fifty healthy lung images from Kaggle repository “Chest X-Ray Images (Pneumonia)” [ 14 ] . The ResNet-50 obtained the best result achieving an accuracy of 98%. Hemdan et al. [ 15 ] proposed a COVIDX-Net composed of seven popular CNN models and used the same dataset considered by [ 12 ] and achieved 90% in terms of accuracy. However, only 25 samples of COVID-19 positive and 25 samples of negative images were considered. Sethy and Behera [ 16 ] also considered the same dataset of [ 15 ] . Their study states that the ResNet-50 as a feature extractor and Support Vector Machine (SVM) classifier provided the best performance obtained an accuracy of 95.38%. Wang and Wong [ 17 ] proposed a COVID-Net architecture, an open-source CNN created to detect COVID-19 on chest X-ray images. The authors used a dataset created exclusively to support COVID-Net experimentation, which obtained 93.3% accuracy in classifying normal, non-COVID pneumonia, and COVID-19 classes. Apostolopoulos and Mpesiana [ 18 ] adopted different pre-trained network architectures to address the task of classification of COVID-19 in chest X-ray images and achieved 96.78% of accuracy with MobileNet v2 model. Khan et al. [ 19 ] proposed the CoroNet deep CNN with 71 layers, inspired by Xception (Extreme Inception) and trained on the ImageNet dataset [ 20 ] . According to the authors, the CoroNet was evaluated using a dataset not publicly available for download and achieved an average accuracy of 89.60% for the COVID-19 identification. Ozturk et al. [ 21 ] proposed an approach for early detection of COVID-19 cases using the DarkNet model as a classifier YOLO object detection system and obtained an accuracy of 98.08% for binary classes and 87.02% for multi-class cases. Ucar and Korkmaz [ 22 ] proposed a method based on SqueezeNet [ 23 ] architecture with Bayes optimization and achieved 98.30% of accuracy. Pereira et

Read the complete text