Automatic COVID-19 detection from X-ray images using ensemble learning with convolutional neural network

Das, Amit Kumar; Ghosh, Sayantani; Thunder, Samiruddin; Dutta, Rohit; Agarwal, Sachin; Chakrabarti, Amlan

doi:10.1007/s10044-021-00970-4

Cited by 165 publications

(97 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SqueezNet turned out to be the best model with a sensitivity of 98% and a specificity of 92.9%. Das et al [17] have developed a new model with a weighted average ensembling method; the model comprises of three pre-trained CNN models-DenseNet201, Resnet50V2, and InceptionV3. This approach has achieved an accuracy of 95.7% and a sensitivity of 98% in the classification of positive and negative COVID-19 cases.…”

Section: Related Workmentioning

confidence: 99%

A comparative study of multiple neural network for detection of COVID-19 on chest X-ray

Anis

Xuan

Chuah

et al. 2021

EURASIP J. Adv. Signal Process.

View full text Add to dashboard Cite

Coronavirus disease of 2019 or COVID-19 is a rapidly spreading viral infection that has affected millions all over the world. With its rapid spread and increasing numbers, it is becoming overwhelming for the healthcare workers to rapidly diagnose the condition and contain it from spreading. Hence it has become a necessity to automate the diagnostic procedure. This will improve the work efficiency as well as keep the healthcare workers safe from getting exposed to the virus. Medical image analysis is one of the rising research areas that can tackle this issue with higher accuracy. This paper conducts a comparative study of the use of the recent deep learning models (VGG16, VGG19, DenseNet121, Inception-ResNet-V2, InceptionV3, Resnet50, and Xception) to deal with the detection and classification of coronavirus pneumonia from pneumonia cases. This study uses 7165 chest X-ray images of COVID-19 (1536) and pneumonia (5629) patients. Confusion metrics and performance metrics were used to analyze each model. Results show DenseNet121 (99.48% of accuracy) showed better performance when compared with the other models in this study.

show abstract

Section: Related Workmentioning

confidence: 99%

A comparative study of multiple neural network for detection of COVID-19 on chest X-ray

Anis

Xuan

Chuah

et al. 2021

EURASIP J. Adv. Signal Process.

View full text Add to dashboard Cite

show abstract

“… [24] Light weight DNN 145+145+145 98.7 98.3 Das et al. [26] Ensemble learning + CNN 438 +_+ 333 91.62 X Rahman et al. [28] U-Net+CNN 3616+6012+8851 X 96.29 Shakarami et al.…”

Section: Resultsmentioning

confidence: 99%

“…To expand the size of the small set of COVID-19 images, one of the promising methods is data augmentation [11] , [13] . Continuous efforts are made employing various machine learning (ML) algorithms where the database of increased number of images are used [21] , [22] , [23] , [24] , [25] , [26] , [27] , [28] , [29] . Chandra et al.…”

Section: Introductionmentioning

confidence: 99%

An optimized KELM approach for the diagnosis of COVID-19 from 2D-SSA reconstructed CXR Images

Mohanty

Dora

2021

Optik

View full text Add to dashboard Cite

The COVID-19 is proved to be the most infectious disease of the current century with a high mortality rate world-wide. The current RT-PCR test standard for the diagnosis of COVID-19 is an invasive and time-consuming procedure, whereas the chest X-ray (CXR) images provide a non-invasive and time/cost-effective method for COVID-19 diagnosis. The current existing deep learning methods for the detection and diagnosis of CXR images provide biased results for the small size dataset available. Hence, in the present work, a conventional yet efficient method is proposed classifying the CXR images into COVID-19 , Pneumonia , and Normal . The proposed approach pre-processes the CXR images using 2D singular spectrum analysis (SSA) for image reconstruction which enhances the feature inputs to the classifier. The features are extracted from the reconstructed images using a block-based GLCM approach. Then, a grasshopper-based Kernel extreme learning machine (KELM) is proposed which finds the optimal features and kernel parameters of KELM at the same instance. From the experimental analysis, it is seen that the present work outperforms that of other competent schemes in terms of classification accuracy with a minimal set of features extracted from the first 2 eigen components of the 2D-SSA reconstructed image with 5 × 5 decomposition.

show abstract

“…This approach improved the pre-training of the model, and it was effective for dealing with irregularities in the dataset. Amit Kumar das [8] applied collaborative approach on deep learning models, and they used ResNet50 V2, DenseNet201, and Inception V3 as the weak learners. Dataset was classified into two classes Covid Positive and Covid Negative, with 538 and 468 images, respectively.…”

Section: B Deep Learningmentioning

confidence: 99%

Performance Analysis of Deep Learning Frameworks for COVID 19 Detection

Naviwala

Qureshi

2021

2021 International Conference on Digital Futures and Transformative Technologies (ICoDT2)

View full text Add to dashboard Cite

caused by disease, can be observed with the help of chest Xray image. Due to these unique features, COVID-19 can be detected by training deep learning models.Ghosh et al.[8] trained a DenseNet201, ResNet50V2, and Inceptionv3 model with COVID-19 and non-COVID-19 patients chest X Rays, and then the results of these three models were ensembled to get improved results. Abbas et al. [9] applied decompose, transfer and compose (DeTraC) approach. DeTraC model was able to achieve 93% accuracy. Ozturk et al.[10] proposed one for binary classification (Covid and no covid) and the other for multiclass classification (Covid, non-covid, and pneumonia). Minaee et al. [11] dataset has 5184 CXR, out of which 5000 images were for the non-covid-19 patient and 187 images were for Covid 19 patient. They fined tune the last layer of pre-trained ResNet18, ResNet50, SqueezeNet, and DenseNet deep convolution network models with CXR images.In this paper, four famous deep learning models VGG 16, Resnet18, Inception V3, and AlexNet deep convolution model are used on two datasets. Due to the limited data, all the models were pre-trained, and then the last layer of the network was fined tuned with COVID-19 patient chest X-Ray images. Another experiment was performed on ResNet18 in which the model was trained on the COVID-Xray-5k dataset but it was tested with the COVID-19 Radiographer dataset [12]. Models were training on the epochs size of 24, but as an experiment, the number of epochs was increased to 30 and ResNet18 was retrained. Overall all the models perform well with the AUC value between 0.985 -0.995. The rest of the paper is organized as follows; In the second section literature review will be discussed. Section three will focus on the methodology, and in section four, we define the result of the study.

show abstract

Automatic COVID-19 detection from X-ray images using ensemble learning with convolutional neural network

Cited by 165 publications

References 32 publications

A comparative study of multiple neural network for detection of COVID-19 on chest X-ray

A comparative study of multiple neural network for detection of COVID-19 on chest X-ray

An optimized KELM approach for the diagnosis of COVID-19 from 2D-SSA reconstructed CXR Images

Performance Analysis of Deep Learning Frameworks for COVID 19 Detection

Contact Info

Product

Resources

About