Automated Detection of COVID-19 from CT Scans using Convolutional

Neural Networks

Rohit Lokwani, Ashrika Gaikwad, Viraj Kulkarni, Anirudha Pant and Amit Kharat

DeepTek Inc., Pune, India

Keywords:

Artiﬁcial Intelligence, Machine Learning, Neural Networks, Deep Learning, Medical Imaging Analysis,

COVID-19, Radiology.

Abstract:

COVID-19 is an infectious disease that causes respiratory problems similar to those caused by SARS-CoV

(2003). In this paper, we propose a prospective screening tool wherein we use chest CT scans to diagnose the

patients for COVID-19 pneumonia. We use a set of open-source images, available as individual CT slices,

and full CT scans from a private Indian Hospital to train our model. We build a 2D segmentation model using

the U-Net architecture, which gives the output by marking out the region of infection. Our model achieves a

sensitivity of 0.96 (95% CI: 0.88-1.00) and a speciﬁcity of 0.88 (95% CI: 0.82-0.94). Additionally, we derive

a logic for converting our slice-level predictions to scan-level, which helps us reduce the false positives.

1 INTRODUCTION

Coronaviruses are a large family of RNA viruses that

are usually known to cause respiratory tract illnesses

like the common cold. They appear crown-like due to

their spiked surface and are categorized into 4 ma-

jor groups: alpha, beta, gamma, and delta. Most

coronaviruses affect animals and can be transmit-

ted between animals and humans (Kong and Agar-

wal, 2020). COVID-19 is the latest addition to the

list of animal-to-human transmissions, preceded by

SARS and MERS. COVID-19 is an infectious dis-

ease that has affected more than 6.8 million people

in the world as of June 8, 2020. The most com-

mon clinical manifestations include fever (83% of

patients), cough (82% of patients), and shortness of

breath (31% of patients) (Chen et al., 2020b). The

hallmarks of COVID-19 include bilateral distribution

of minute patchy shadows and ground-glass opacity

in the nascent stages. The progression of this dis-

ease is marked by the spread of these opacities and

inﬁltrates to both the lungs (Wang et al., 2020b).

The World Health Organization has published sev-

eral testing protocols for detecting the disease (Wang

et al., 2020a). The most commonly used reference

test for the diagnosis of COVID-19 is the real-time

reverse transcription-polymerase chain reaction (RT-

PCR) (Gundlapally et al., ).

Reverse Transcription Polymerase Chain Reaction

(RT-PCR) test is the key approach used for diagnos-

ing COVID-19. However, it has some limitations;

their shortcomings include the complex process used

for specimen collection, the amount of time required

for the analysis, and variability in the accuracy of the

tests (Bullock et al., 2020). Apart from this, a ma-

jor hurdle in controlling the spread of the disease is

the accuracy and shortage of testing kits (Zhao et al.,

2020a). Hence, computer-based detection assisted by

an expert in the loop with minimal infrastructure is

proposed as an alternative to testing kits and vaccines.

Computer-aided detection has helped in detecting, lo-

calizing, and segmenting out a varied set of diseases

using medical imaging analysis. In particular, ma-

chine learning is being used for medical imaging anal-

ysis by developing deep-learning systems that extract

the spatio-temporal representative features from an

image, analyze them, and decide the diagnostic out-

comes (Wang et al., 2020b).

The most common, economical, and easy-to-use

medical imaging and diagnostic technique is chest ra-

diography or chest X-rays. This technique plays an

important role in the diagnosis of lung diseases. Ex-

pert radiologists use chest X-ray images (CXRs) to

detect pathologies like pneumonia, tuberculosis, at-

electasis, inﬁltrates, and early lung cancer (Qin et al.,

2018). But, detecting COVID-19 using CXRs is

challenging due to the less evident visual features in

CXRs caused by the overlapping of ribs and soft tis-

sues and low contrast (Zhang et al., 2020). The lim-

ited availability of annotated images adds to the dif-

ﬁculty. The RT-PCR test is very speciﬁc but has a

lower sensitivity of 65-95%, which means that the

test can be negative even when the patient is infected

(Fang et al., 2020)(Ai et al., 2020). These short-

Lokwani, R., Gaikwad, A., Kulkarni, V., Pant, A. and Kharat, A.

Automated Detection of COVID-19 from CT Scans using Convolutional Neural Networks.

DOI: 10.5220/0010293605650570

In Proceedings of the 10th International Conference on Pattern Recognition Applications and Methods (ICPRAM 2021), pages 565-570

ISBN: 978-989-758-486-2

565

comings can be resolved by using chest CT scans,

a cross-sectional imaging modality with high accu-

racy and speed, instead of CXRs. A recent study of

the coronavirus infection on the cruise ship “Diamond

Princess” showed evidence of the lung parenchymal

pattern (classic for COVID-19) on CT studies of the

chest in 54% of the asymptomatic cases (Inui et al.,

2020).

Most of the recent literature reported that COVID-

19-positive patients had characteristic features highly

evident in the CT scan images (Xie et al., 2020).

These features included different degrees of ground-

glass opacities with or without crazy-paving sign,

multifocal organizing pneumonia, and architectural

distortion in a peripheral distribution (Ai et al., 2020).

COVID-19 eventually develops into chronic pneumo-

nia, and thus the visual symptoms it has are similar to

those of bacterial and viral pneumonia. In CT scans,

the ground-glass opacities are more similar to con-

solidation (Wang et al., 2020b). Studies have proven

that chest CT has a higher sensitivity for the diagnosis

of COVID-19 as compared with RT-PCR tests taken

from swab samples (Ai et al., 2020). To curb human-

to-human transmission and isolate the affected from

the healthy, it is essential to detect the presence of

COVID-19 at an early stage. This is where CT assists

in the detection of minor infections (Anthimopoulos

et al., 2016).

In this paper, we propose a prospective technique

based on artiﬁcial neural networks wherein our model

predicts the CT scan as COVID-19 positive or nega-

tive. This screening tool can help prioritize the treat-

ment for patients with COVID-19 visual manifesta-

tions in their CT scans.

2 RELATED WORK

In the past few years, deep learning has evolved

as a technique with its capabilities extending from

classiﬁcation and object detection to segmentation

in medical image analysis. Some studies showed

better results than expert radiologists. Rajpurkar

et al. (Rajpurkar et al., 2017) proposed and pre-

sented a DenseNET-121 model for pneumonia detec-

tion which performed binary classiﬁcation on CXRs

using CNNs. This paper used F-1 score as the primary

metric but failed to specify the prevalence of the set.

Qin et al. (Qin et al., 2018) proposed pneumonia and

pulmonary edema classiﬁcation by extracting textural

features. Parveen et al. (Parveen and Sathik, 2011)

used an FCM clustering algorithm to detect pneumo-

nia, where they showed that the lung area of the chest

appeared like a black or dark gray shaded region when

it became infected with pneumonia.

Recently, there have been many developments

in detecting COVID-19 from CXRs and CTs. Xu

et al. (Xu et al., ) proposed a 3D deep learning

model that categorized CT scans as either COVID-

19 pneumonia-positive or viral pneumonia-positive.

They trained a location-attention classiﬁcation model

and used the predicted probabilities to give a predic-

tion calculated by a Bayesian function. Their best

model gave a recall of 86.7% and it needed further

validation on multi-clinical studies. Chen et al. (Chen

et al., 2020a) built a model using UNet++ (Zhou

et al., 2018), a powerful architecture for medical im-

age segmentation, and used a 3-consecutive slice and

quadrant-based post-processing approach to mark a

scan as positive or negative. This post-processing

approach helped them reduce the number of false

positives. Several studies have addressed diagnosis

as a binary classiﬁcation problem, i.e., healthy vs.

COVID-19-positive (Bullock et al., 2020). For exam-

ple, Wang et al. (Wang et al., 2020b) used a mod-

iﬁed Inception neural network architecture and at-

tained an accuracy of 79.3%. This model was trained

on the CTs having severe pathological infections and

it hence needs to be tested for all pathological stages

to validate this in real-world scenarios. Shan et al.

(Shan et al., 2020) developed a deep learning system

that automatically quantiﬁed infection regions of in-

terest (ROIs) and their volumetric ratios with respect

to the lung. Li et al. (Li et al., 2020) put forth a 3D

deep learning model, where they combined the 2D lo-

cal and 3D global features using a max-pooling opera-

tion and predicted the class using the probability score

from the softmax activation. One of the signiﬁcant

limitations of this paper is that the model seems to

give 90% sensitivity on their test set, but hasn’t been

tested on the out-of-sample test set. In our paper, we

overcome this limitation. Jianpeng et al. (Zhang et al.,

2020) proposed a deep-learning architecture to differ-

entiate COVID-19 cases from non-COVID-19 cases

from CXRs. Their model is composed of three com-

ponents: a backbone network, a classiﬁcation head,

and an anomaly detection head. The backbone net-

work extracts the high-level features and feeds them

to the rest of the heads. Zhou et al. (Zhou et al., 2020)

use a U-Net with attention mechanism to utilize rich

contextual features from the U-Net encoder, they train

their model using focal tversky loss for small lesion

segmentation. The model has been trained and eval-

uated on a small dataset of around 829 slices. Yan

et al. (Yan et al., 2020) propose a Feature Varia-

tion block which enhances the global intensity of the

pixels around the lung region and uses Progressive

Atrous Spatial Pyramid Pooling to handle manifes-

ICPRAM 2021 - 10th International Conference on Pattern Recognition Applications and Methods

566

tations at various scales. This model achieves state-

of-the-art performance on 3D U-Net models attaining

dice coefﬁcient of 0.726 on datasets from China and

Germany.

Several other approaches used a 3-category classi-

ﬁcation method, differentiating healthy patients from

pneumonia and COVID-19. Xu et al. (Xu et al.,

) used classical ResNet architectures, adding fully-

connected layers at the end, and took the classiﬁca-

tion approach to solve the problem. He et al. (He

et al., 2016) used ResNets for feature extraction, and

Song et al. (Song et al., 2020) used the Feature Pyra-

mid Networks (Lin et al., 2017), which are the back-

bone in U-Nets, for learning ﬁne-grained features in

the images.

Gurujit et al. (Randhawa et al., 2020) identiﬁed

an intrinsic COVID-19 genomic signature and used

it together with a machine learning-based alignment-

free approach for an ultra-fast, scalable, and highly

accurate classiﬁcation of whole COVID-19 genomes.

3 DATA

We used COVID-19-positive and non-COVID data

from GitHub (Zhao et al., 2020b) and consolidation

and healthy CT scans from a private Indian hospital.

The data obtained contained 275 CT scans labeled as

COVID-19-positive. The ground truth in these im-

ages was decided on the basis of their RT-PCR test

results. These CT images had different sizes from

143 patient cases (Zhao et al., 2020a). The scans

differed in voxel sizes but had the same aspect ra-

tio. In total, the data contained 5212 slices and was

split at patient-level into training, validation, and test

sets. Each set had a prevalence of 20% of positive

cases. As the available open-source data had resolu-

tions varying from 256x256 to 768x768, we resize the

input to 512x512 pixels. This input size was the me-

dian size of the images in the dataset and sufﬁced our

computational requirements. This model was trained

using a GPU with 16GB RAM. The original images

were in the unsigned int8 format, in the range of [0,

255]. We converted these images to ﬂoating-point 16,

in the range of [0, 1]. The output masks were in the

binary form [0, 1] at pixel-level, where 1s indicated

the region of interest. Table 1 shows the detailed dis-

tribution of data.

Qualiﬁed radiologists inspected the CT slices one-

by-one and classiﬁed each slice into one of two

classes: COVID and NON-COVID. The COVID class

contained slices where typical ﬁndings including bi-

lateral pulmonary parenchymal ground-glass and con-

solidative pulmonary opacities, sometimes with a

Table 1: Slice-Level Dataset splits.

Dataset COVID-

NON-

COVID

Total

slices

Training 657 2628 3285

Validation 120 477 597

Test 266 1064 1330

rounded morphology and a peripheral lung distribu-

tion (Chung et al., 2020) were observed. Ground-

glass opaciﬁcation was deﬁned as hazy increased lung

attenuation with preservation of bronchial and vascu-

lar margins, and consolidation was deﬁned as opaciﬁ-

cation with obscuration of margins of vessels and air-

way walls (Hansell et al., 2008). Notably, lung cavita-

tion, discrete pulmonary nodules, and lymphadenopa-

thy were marked as negative. Other slices where

above manifestations were not seen were marked as

NON-COVID. For COVID slices, the radiologists

also highlighted the region of interest where the man-

ifestations were observed using polygon masks.

Figure 1 shows an example of the annotation. We

used positive slices from COVID positive CTs and

negative slices from healthy scans for training our

model, this helped us overcome the challenge of using

any false negatives to train the model. All the images

in Table 1. were annotated by the radiologists, con-

verted to masks and used as ground truth for training

and evaluating the model.

Figure 1: (Left) original image and (right) annotated ROI.

4 METHODOLOGY

In this section, we give a brief overview of our train-

ing and the inference algorithms.

We used U-Net (Ronneberger et al., 2015) for

medical image segmentation, which uses the concept

of deconvolution (Zeiler and Fergus, 2014). U-Nets

are built on the architecture of fully convolutional net-

works. The most important property of U-Net is the

shortcut connections between the layers of equal reso-

lution in the encoder path and the decoder path. These

connections provide essential high-resolution features

to the deconvolution layers (Hesamian et al., 2019).

Automated Detection of COVID-19 from CT Scans using Convolutional Neural Networks

567

Here, we used Xception (Chollet, 2017) as the en-

coder for U-Net. Xception with its depthwise separa-

ble convolutions and residual connections, has proven

to give better performance as compared to other mod-

els with similar parameters (Chollet, 2017).

Initially, we used ImageNet weights to train the

model, but the model predicted a cluster of pixels

instead of coherent masks. As we did not have a

CT model for the same architecture, we used trans-

fer learning by ﬁne-tuning a network pre-trained on

CXRs for the same problem but a different task (Shin

et al., 2016). Transfer learning tends to give bet-

ter performance when the tasks of source and target

network are more similar; yet even transferring the

weights of far and distant tasks has been proved to

be better than random initialization (Yosinski et al.,

2014).

Here, we have tried to solve the problem of dis-

tinguishing COVID-19 cases from non-COVID-19 by

using weights from our COVID-19 vs healthy model,

as pre-trained weights for this model already gave a

sensitivity of 0.9 with a speciﬁcity of 0.8. We then

built a CT model for consolidation vs healthy and

later ﬁne-tuned our model for COVID-19 vs non-

COVID-19.

In the training stage, we use binary cross entropy

as the loss function and the standard Adaptive Adam

Optimizer with a batch size of 4. We set the maxi-

mum epochs to 50 and set the learning rate to 10

−4

which is decayed on the plateau after patience of 4

epochs. We resize each training image to a ﬁxed size

of 512× 512 pixels. To alleviate the overﬁtting of our

model on the training data from a particular source,

we try to include data from varied sources. One of

the drawbacks of having a 2D CT model is that the

inference tends to be slow. Since our model has a

sensitivity of 0.964, we plan to use speciﬁc slices for

inference.

5 RESULTS

We tested our model using varied sets of data from

different sources. We initially evaluated the model

on our test set, consisting of 1330 images, in which

COVID-19-positive samples had a prevalence of

20%. Our model gave a sensitivity of 0.963 (95% CI:

0.94-0.98) and a speciﬁcity of 0.936 (95% CI: 0.92-

0.95). The dice coefﬁcient on positive samples was

0.561. Figures 2 and 3 show the superimposed masks

on one of the slices.

Apart from this, we evaluated the model on a to-

tal of 140 scans with a prevalence of 20% for positive

cases. These scans were tested on data from three

Figure 2: (Left) original image and (right) corresponding

predicted mask.

Figure 3: (Left) original image and (right) corresponding

predicted mask.

sources. One source contained scans from Italy and

China, while the remaining came from two separate

private Indian hospitals. After passing these images

through our model, we sorted the slices as per the

position of the slice in the CT scan. We observed a

pattern wherein the consecutive slices had the same

predictions, which is expected from a radiology per-

spective. Figure 4 provides an example of the predic-

tions for a positive CT scan. Here we see the expected

pattern of consecutive slices, predicted as positive by

the model.

Hence, we convert the slice-level prediction to

scan-level prediction using the logic that if 15 consec-

utive slices in a scan are marked as positive, then we

mark the scan as positive (Chen et al., 2020a). Table

2 shows the results obtained at scan-level.

Table 2: Scan-Level performance of the model on the test

set.

Performance Metric Value 95% C.I.

Sensitivity 0.964 (0.88,1)

Speciﬁcity 0.884 (0.82,0.94)

F1-score 0.794 (0.68,0.89)

6 DISCUSSION

The diagnosis of COVID-19 using CXRs and CT

scans has gained signiﬁcance since the ubiquitous

ICPRAM 2021 - 10th International Conference on Pattern Recognition Applications and Methods

568

Figure 4: (Left) original image and (right) corresponding

predicted mask.

spread of this disease. But, chest CT scans usually

tend to show the region of infection more clearly than

CXRs (Kong and Agarwal, 2020). A limitation of

this study is that the patterns considered for COVID-

19 were few in number– notably consolidation and

ground glass opacity. These patterns might vary re-

gionally where pleural effusion could be observed in

COVID-19 infected patients. These patterns can even

overlap with other pathology manifestations.

Another limitation is we have not considered the

clinical history of the patient. The real-world util-

ity of this tool can be enhanced once it considers the

radiological and clinical parameters to determine the

ultimate outcome. Our current implementation is a

2D model built at slice-level. Since a CT study could

have the number of slices running into thousands, this

2D model certainly adds to the time complexity of

processing the whole scan. Although we are satis-

ﬁed with the performance our model currently shows

on the data from diverse distributions, deploying the

model in production is a challenge, given the time

complexity.

In the future, we plan to implement a 3D model

that will take the whole CT scan as input and give out

masks for the infected areas. The primary challenge

with this approach will be the requirement of a lot

of annotated data to give an equivalent performance.

Additionally, we propose a model that differentiates

between COVID-19 and chronic and viral pneumo-

nia and address the challenges associated with it, like

ﬁne-grained, accurate annotations and large amounts

of data for all the speciﬁed categories. In conclu-

sion, chest CT has proved to have a higher sensitiv-

ity than RT-PCR tests (Ai et al., 2020). Our analysis

suggests that chest CT can be a potential alternative

for COVID-19 screening and evaluation, especially

in epidemic situations where the spread is uncontrol-

lable, and diagnosis needs to be done with celerity.

REFERENCES

Ai, T., Yang, Z., Hou, H., Zhan, C., Chen, C., Lv, W., Tao,

Q., Sun, Z., and Xia, L. (2020). Correlation of chest ct

and rt-pcr testing in coronavirus disease 2019 (covid-

19) in china: a report of 1014 cases. Radiology, page

200642.

Anthimopoulos, M., Christodoulidis, S., Ebner, L., Christe,

A., and Mougiakakou, S. (2016). Lung pattern classi-

ﬁcation for interstitial lung diseases using a deep con-

volutional neural network. IEEE transactions on med-

ical imaging, 35(5):1207–1216.

Bullock, J., Pham, K. H., Lam, C. S. N., Luengo-Oroz, M.,

et al. (2020). Mapping the landscape of artiﬁcial intel-

ligence applications against covid-19. arXiv preprint

arXiv:2003.11336.

Chen, J., Wu, L., Zhang, J., Zhang, L., Gong, D., Zhao, Y.,

Hu, S., Wang, Y., Hu, X., Zheng, B., et al. (2020a).

Deep learning-based model for detecting 2019 novel

coronavirus pneumonia on high-resolution computed

tomography: a prospective study. medRxiv.

Chen, N., Zhou, M., Dong, X., Qu, J., Gong, F., Han, Y.,

Qiu, Y., Wang, J., Liu, Y., Wei, Y., et al. (2020b). Epi-

demiological and clinical characteristics of 99 cases of

2019 novel coronavirus pneumonia in wuhan, china: a

descriptive study. The Lancet, 395(10223):507–513.

Chollet, F. (2017). Xception: Deep learning with depthwise

separable convolutions. In Proceedings of the IEEE

conference on computer vision and pattern recogni-

tion, pages 1251–1258.

Chung, M., Bernheim, A., Mei, X., Zhang, N., Huang, M.,

Zeng, X., Cui, J., Xu, W., Yang, Y., Fayad, Z. A., et al.

(2020). Ct imaging features of 2019 novel coronavirus

(2019-ncov). Radiology, 295(1):202–207.

Fang, Y., Zhang, H., Xie, J., Lin, M., Ying, L., Pang, P.,

and Ji, W. (2020). Sensitivity of chest ct for covid-19:

comparison to rt-pcr. Radiology, page 200432.

Gundlapally, P., Pingili, S., and Doragolla, B. A novel coro-

navirus disease (covid-19) outbreak as a pandemic cri-

sis.

Hansell, D. M., Bankier, A. A., MacMahon, H., McLoud,

T. C., Muller, N. L., and Remy, J. (2008). Fleischner

society: glossary of terms for thoracic imaging. Radi-

ology, 246(3):697–722.

He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep resid-

ual learning for image recognition. In Proceedings of

the IEEE conference on computer vision and pattern

recognition, pages 770–778.

Hesamian, M. H., Jia, W., He, X., and Kennedy, P. (2019).

Deep learning techniques for medical image segmen-

tation: Achievements and challenges. Journal of dig-

ital imaging, 32(4):582–596.

Inui, S., Fujikawa, A., Jitsu, M., Kunishima, N., Watan-

abe, S., Suzuki, Y., Umeda, S., and Uwabe, Y.

(2020). Chest ct ﬁndings in cases from the cruise

ship “diamond princess” with coronavirus disease

2019 (covid-19). Radiology: Cardiothoracic Imag-

ing, 2(2):e200110.

Kong, W. and Agarwal, P. P. (2020). Chest imaging ap-

pearance of covid-19 infection. Radiology: Cardio-

thoracic Imaging, 2(1):e200028.

Automated Detection of COVID-19 from CT Scans using Convolutional Neural Networks

569

Li, L., Qin, L., Xu, Z., Yin, Y., Wang, X., Kong, B., Bai,

J., Lu, Y., Fang, Z., Song, Q., et al. (2020). Artiﬁ-

cial intelligence distinguishes covid-19 from commu-

nity acquired pneumonia on chest ct. Radiology, page

200905.

Lin, T.-Y., Doll

ar, P., Girshick, R., He, K., Hariharan, B.,

and Belongie, S. (2017). Feature pyramid networks

for object detection. In Proceedings of the IEEE con-

ference on computer vision and pattern recognition,

pages 2117–2125.

Parveen, N. and Sathik, M. M. (2011). Detection of pneu-

monia in chest x-ray images. Journal of X-ray Science

and Technology, 19(4):423–428.

Qin, C., Yao, D., Shi, Y., and Song, Z. (2018). Computer-

aided detection in chest radiography based on artiﬁcial

intelligence: a survey. Biomedical engineering online,

17(1):113.

Rajpurkar, P., Irvin, J., Zhu, K., Yang, B., Mehta, H., Duan,

T., Ding, D., Bagul, A., Langlotz, C., Shpanskaya, K.,

et al. (2017). Chexnet: Radiologist-level pneumonia

detection on chest x-rays with deep learning. arXiv

preprint arXiv:1711.05225.

Randhawa, G. S., Soltysiak, M. P., El Roz, H., de Souza,

C. P., Hill, K. A., and Kari, L. (2020). Machine learn-

ing using intrinsic genomic signatures for rapid classi-

ﬁcation of novel pathogens: Covid-19 case study. Plos

one, 15(4):e0232391.

Ronneberger, O., Fischer, P., and Brox, T. (2015). U-net:

Convolutional networks for biomedical image seg-

mentation. In International Conference on Medical

image computing and computer-assisted intervention,

pages 234–241. Springer.

Shan, F., Gao, Y., Wang, J., Shi, W., Shi, N., Han, M., Xue,

Z., and Shi, Y. (2020). Lung infection quantiﬁcation

of covid-19 in ct images with deep learning. arXiv

preprint arXiv:2003.04655.

Shin, H.-C., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues,

I., Yao, J., Mollura, D., and Summers, R. M. (2016).

Deep convolutional neural networks for computer-

aided detection: Cnn architectures, dataset charac-

teristics and transfer learning. IEEE transactions on

medical imaging, 35(5):1285–1298.

Song, Y., Zheng, S., Li, L., Zhang, X., Zhang, X., Huang,

Z., Chen, J., Zhao, H., Jie, Y., Wang, R., et al. (2020).

Deep learning enables accurate diagnosis of novel

coronavirus (covid-19) with ct images. medRxiv.

Wang, M., Zhou, Y., Zong, Z., Liang, Z., Cao, Y., Tang,

H., Song, B., Huang, Z., Kang, Y., Feng, P., et al.

(2020a). A precision medicine approach to managing

2019 novel coronavirus pneumonia. Precision Clini-

cal Medicine, 3(1):14–21.

Wang, S., Kang, B., Ma, J., Zeng, X., Xiao, M., Guo, J.,

Cai, M., Yang, J., Li, Y., Meng, X., et al. (2020b). A

deep learning algorithm using ct images to screen for

corona virus disease (covid-19). MedRxiv.

Xie, X., Zhong, Z., Zhao, W., Zheng, C., Wang, F., and Liu,

J. (2020). Chest ct for typical 2019-ncov pneumonia:

relationship to negative rt-pcr testing. Radiology, page

200343.

Xu, X., Jiang, X., Ma, C., Du, P., Li, X., Lv, S., Yu, L.,

Chen, Y., Su, J., Lang, G., et al. Deep learning system

to screen coronavirus disease 2019 pneumonia. arxiv

2020. arXiv preprint arXiv:2002.09334.

Yan, Q., Wang, B., Gong, D., Luo, C., Zhao, W., Shen,

J., Shi, Q., Jin, S., Zhang, L., and You, Z. (2020).

Covid-19 chest ct image segmentation–a deep con-

volutional neural network solution. arXiv preprint

arXiv:2004.10987.

Yosinski, J., Clune, J., Bengio, Y., and Lipson, H. (2014).

How transferable are features in deep neural net-

works? In Advances in neural information processing

systems, pages 3320–3328.

Zeiler, M. D. and Fergus, R. (2014). Visualizing and under-

standing convolutional networks. In European confer-

ence on computer vision, pages 818–833. Springer.

Zhang, J., Xie, Y., Li, Y., Shen, C., and Xia, Y. (2020).

Covid-19 screening on chest x-ray images using deep

learning based anomaly detection. arXiv preprint

arXiv:2003.12338.

Zhao, J., Zhang, Y., He, X., and Xie, P. (2020a). Covid-

ct-dataset: a ct scan dataset about covid-19. arXiv

preprint arXiv:2003.13865.

Zhao, J., Zhang, Y., He, X., and Xie, P. (2020b). Covid-

ct-dataset: a ct scan dataset about covid-19. arXiv

preprint arXiv:2003.13865.

Zhou, T., Canu, S., and Ruan, S. (2020). An automatic

covid-19 ct segmentation based on u-net with atten-

tion mechanism. arXiv preprint arXiv:2004.06673.

Zhou, Z., Siddiquee, M. M. R., Tajbakhsh, N., and Liang, J.

(2018). Unet++: A nested u-net architecture for medi-

cal image segmentation. In Deep Learning in Medical

Image Analysis and Multimodal Learning for Clinical

Decision Support, pages 3–11. Springer.

ICPRAM 2021 - 10th International Conference on Pattern Recognition Applications and Methods

570