Comparing the performance of ResNets on COVID-19 diagnosis using CT scans

Abstract

To the date, the Coronavirus disease 2019 (COVID-19) is still spreading around the world. Reverse transcription polymerase chain reaction (RT-PCR), which is regarded as the gold standard by the anti-epidemic front. Unfortunately, due to its lack of sensitivity (only 60%~71%), unstability, long processing time, some alternative diagnostic tools including computed tomography (CT) have been developed using convolutional neural networks (CNNs). This paper retrospectively collected 3227 CT scans (including 1601 COVID-19 positive and 1626 negative samples), on which ResNet-18 outperforms all three classical ResNets with the precision of 0.971, accuracy of 0.943, F1 of 0.942, sensitivity of 0.914, specificity of 0.973 and AUC of 0.985 respectively. Finally, the visualization technique classification activation map (CAM) was applied for lesions localization.

A hybrid CNN–KNN approach for identification of COVID-19 with 5-fold cross validation

“…Several papers [ 13 , 14 , 16 , [39] , [40] , [41] , [42] ] experimented on the same dataset to detect COVID-19 patients. To compare the result obtained from the proposed hybrid method with previous studies, those who applied the deep learning method on CT scan images are listed in Table .12 .…”

“…Different deep learning techniques are used in Refs. [ 13 , 14 , 16 , 39 , 40 ] for COVID-19 classifications. Transfer learning is applied on [ 13 , 14 , 16 , 38 ] and feature extraction is manipulated by pre-trained GoogleNet in Ref.…”

“…Cai [ 13 ] has combined two publicly available datasets [ 14 , 15 ] and trained them on ResNet architecture. Instead of having a decent number of images in the dataset, it shows an accuracy of 94.3%.…”

COVID-19 Identification in CLAHE Enhanced CT Scans with Class Imbalance using Ensembled ResNets

“…With the spread of the disease to over 214 countries [11] of the world with widely varying severity and the low detection rates of the disease due to its asymptomatic nature, the identification of the disease is proving to be a challenge. [12] follow the architecture of Convolutional Neural Networks (or CNNs) [13]. They exhibit high control over the information that flows from one cell to the next in the cell using a parameterized forget gate.…”

A Literature Review: Detection of COVID-19 in Computed Tomography Images Using Deep Learning

“…Por fim temos os autores que fizeram uma revisão dos modelos mais difundidos da literatura, como os autores [Seum et al 2020], que apresentam uma avaliac ¸ão de diferentes CNN pré-treinadas, com destaque para ResNet18 e DenseNet121 entre os mo-delos com e sem segmentac ¸ão, respectivamente. Em [Cai et al 2020] é realizado um estudo de três tipos de ResNet, onde a ResNet-18 obteve os melhores resultados. Os autores em [Yener and Oktay 2020] trazem um estudo sobre três modelos populares de CNN usando aprendizagem por transferência, onde a VGG16 e VGG19 obtiveram resultados semelhantes e melhores que a Xception.…”