PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

The novel coronavirus (COVID-19) pandemic is pressurizing the healthcare systems across the globe and few of them are on the verge of failing. The detection of this virus as early as possible will help in contaminating the spread of it as the virus is mutating itself as fast as possible and currently there are about 4,300 strains of the virus according to the reports. Clinical studies have shown that most of the COVID-19 patients suffer from a lung infection similar to influenza. So, it is possible to diagnose lung infection using imaging techniques. Although a chest computed tomography (CT) scan has been shown to be an effective imaging technique for lung-related disease diagnosis, chest X-ray is more widely available across the hospitals due to its considerably lower cost and faster imaging time than CT scan. The advancements in the area of machine learning and pattern recognition has resulted in intelligent systems that analyze CT Scans or X-ray images and classify between pneumonia and normal patients. This paper proposes KE Sieve Neural Network architecture, which helps in the rapid diagnosis of COVID-19 using chest X-ray images. This architecture is achieving an accuracy of 98.07%. This noninvasive prediction method can assist the doctors in this pandemic and reduce the stress on health care systems.
Content may be subject to copyright.
Precise Prediction of COVID-19 in Chest X-Ray
Images Using KE Sieve Algorithm
S Sai Thejeshwar 1
Hyderabad, India
ssai.thejeshwar.in@ieee.org
Chaitanya Chokkareddy 1
Hyderabad, India
chaitanya.chokkareddy@gmail.com
Dr. K Eswaran 1, 2
Hyderabad, India
kumar.e@gmail.com
Abstract
The novel coronavirus (COVID-19)
pandemic is pressurizing the healthcare systems across
the globe and few of them are on the verge of failing. The
detection of this virus as early as possible will help in
contaminating the spread of it as the virus is mutating
itself as fast as possible and currently there are about
4,300 strains of the virus according to the reports.
Clinical studies have shown that most of the COVID-19
patients suffer from a lung infection similar to influenza.
So, it is possible to diagnose lung infection using imaging
techniques. Although a chest computed tomography
(CT) scan has been shown to be an effective imaging
technique for lung-related disease diagnosis, chest X-ray
is more widely available across the hospitals due to its
considerably lower cost and faster imaging time than CT
scan. The advancements in the area of machine learning
and pattern recognition has resulted in intelligent
systems that analyze CT Scans or X-ray images and
classify between pneumonia and normal patients. This
paper proposes KE Sieve Neural Network architecture,
which helps in the rapid diagnosis of COVID-19 using
chest X-ray images. This architecture is achieving an
accuracy of 98.07%. This noninvasive prediction method
can assist the doctors in this pandemic and reduce the
stress on health care systems.
Keywords— KE Sieve, Non-iterative Training, Image
Processing, Computer Vision, Pattern Recognition,
Machine Learning, COVID-19, X-ray Images.
I. INTRODUCTION
The novel coronavirus (COVID-19), first appeared in
Wuhan, China in December 2019 and has been rapidly
spreading across the countries worldwide [5] [6] and
subsequently threatening the health of billions of humans
and forcing millions of people to stay at home with the
majority of countries declaring lockdowns. It has been
declared as a pandemic by WHO [11]. The rapid escalation
of COVID-19 virus with 2,397,216 confirmed cases and
162,956 confirmed deaths covering 213 countries, areas or
territories, (As on 21 April 2020, 2:00 AM CEST) [12] is
presenting a huge challenge to the governments across the
world, as new cases are being emerged and many new
hotspots are evolving on daily basis.
1 - ALPES.ai
2 - Department of Computer Science and Engineering, Sreenidhi Institute
Of Science and Technology
According to the World Health Organisation (WHO),
“COVID-19 is the infectious disease caused by the most
recently discovered coronavirus. This new virus and disease
were unknown before the outbreak began in Wuhan, China,
in December 2019.” [9]. This COVID-19 is also called the
Severe Acute Respiratory Syndrome CoronaVirus 2
(SARS-CoV-2) [7] [18]. Coronaviruses (CoV) belongs to a
family of viruses that cause colds such as the Severe Acute
Respiratory Syndrome (SARS-CoV) and the Middle East
Respiratory Syndrome (MERS-CoV). This COVID-19 virus
is new to human beings and has not been previously
identified in humankind. It is said that this virus outbreak is
due to the contamination of the virus from animals to
humans [8]. According to studies, [8] [9] the SARS-CoV
virus is contaminated from musk cats to humans, and the
MERS-CoV virus is contaminated from camels to
humans. But, the COVID-19 virus is presumed to be
contaminated from bats to humans [9]. The COVID-19 virus
spreads primarily through droplets of saliva or through
cough and sneezing [9]. This led to the transmission of the
virus from person to person and caused the rapid spread of
the epidemic and had become a pandemic. In this war
against COVID-19, all the countries have united together
and are fighting against the pandemic. But, this is never
about winning at this war, it is about survival. Survival of
mankind on earth.
The common symptoms of the disease include fever,
tiredness, dry cough. Other symptoms include, shortness of
breath, aches and pains, sore throat and very few people had
reported diarrhea, nausea, or a runny nose. [9]
The current diagnostic method for COVID-19 is
Real-Time Polymerase Chain Reaction (RT-PCR) which
uses viral nucleic acid detection in the blood sample.
However, there are a limited number of RT-PCR test kits
available in most of the countries. Therefore, it is necessary
to implement an automatic detection system as a quick
alternative diagnosis option to prevent COVID-19 spreading
among people. [13] [14] [15]
India, the second-most populous country in the world
with 1.2 billion people as of March 2011 [17] and currently
has 18,601 confirmed cases and 590 deaths (As on 21 April
2020, 2:00 AM CEST). [12] Given the size of the
population these numbers may look small. But, it is a big
challenge to the government for the containment of the
spread of the disease. According to the Indian Council of
Medical Research (ICMR) [18] “A total of 4,01,586
1
samples from 3,83,985 individuals have been tested as on 19
April 2020, 9 PM IST. 17,615 individuals have been
confirmed positive among suspected cases and contacts of
known positive cases in India.” Are the number of positive
cases actually low or they are low because of low testing
rate still remains a question. There are many more such
countries with low testing rates [19]. But it is essential to
increase the testing rate to curb the spread of COVID-19.
[15]. Also, some of the eye-opening statistics of India are
that 67 percent of the Indian population resides in rural
areas and 90 percent of medical imaging facilities are in
cities, That makes only 10 percent of imaging facilities
spread across small towns in the country [28].
According to the World Health Organization, COVID-19
also opens holes in the lungs like SARS, giving them a
"honeycomb-like appearance" [10] [16]. So, in this study,
we propose an AI-based pattern recognition system using
the KE Sieve Neural Network model [1] [2] for the
detection of coronavirus infected patients, pneumonia and
healthy patients using chest X-ray radiographs. Also, there
are only 3 radiologists per 1,000,000 people in India [28].
So, an automatic screening system would not only help
India but also countries across the globe to do a rapid
screening and further prevent the spread of the virus. The
novelty of this paper is summarized as follows:
1. The proposed model has an end-to-end system
without manual human intervention [31].
2. Chest X-ray images are the best tool for the
detection of COVID-19.
3. The proposed model classifies between
COVID-19, Healthy, and Pneumonia patients.
4. With an increase in the number of training data in
the future on a day to day basis, the model can
learn as you go.
II. RELATED WORK
There are few studies on the emergence of COVID-19
virus disease. Fang et al. [20] analyzed computed
tomography (CT) scan images using deep neural networks
and reported a sensitivity of 98% using 51 COVID-19
patients. Also, Shi et al. [21] collected a large scale
COVID-19 CT dataset and developed a machine learning
model for COVID-19 screening. Ophir et al. [22] employed
deep learning models to detect COVID-19 on CT images.
Also, Prabira et al. [23] proposed the detection of
COVID-19 using X-ray images based on deep features and
SVM. They have extracted the deep features of CNN
models and fed them to the SVM classifier. They have
obtained 95.38% of accuracy for ResNet50 and Support
Vector Machine (SVM). Fei et al. [24], tried to predict
COVID-19 patients using “VB-Net” neural network to
segment COVID-19 infection regions in CT scans. They
obtained dice similarity coefficients of 91.6%±10.0%. Xu
et al. [25], proposed a classification model that classifies
COVID-19 from viral pneumonia and healthy cases using
pulmonary CT images using deep learning techniques.
Their CNN model has yielded the then highest overall
accuracy of 86.7 % on CT images. Shuai et al. [26], used
CT images to predict COVID-19 cases. They used the
inception transfer-learning model to establish the algorithm.
They obtained an accuracy of 89.5% with a specificity of
88.0% and a sensitivity of 87.0%.
In some analysis, the normal and pneumonia images have
been combined and a binary classification model between
COVID-19 and combined classes has been proposed, which
is not appropriate as the model will then try to ignore the
between-group variance amongst those two classes and the
accuracy thus obtained will not be a true measure. It is a
better approach to have a tri-class classification model.
The drawback of CT imaging is that typically it takes
considerably more time than X-ray imaging. Besides CT
scanners may not be available in many regions, making
timely COVID-19 screening impossible. In contrast,
X-rays are the most common and widely available
diagnostic imaging technique [27] and portable in nature.
Which makes it a better choice than CT imaging-based
detection of COVID-19.
III. PROPOSED MODEL
The KE Sieve Neural Network [1] [1b] [2] is applied to
the Datasets [3] [4] while using features extracted from the
transfer learning of pre-trained model weights on the
VGG-19 [30] CNN model.
A. Architecture
The KE Sieve Algorithm [1] is a non-iterative and
adopts a new approach, which separates N data points of
n-dimension by at least one hyperplane. The number of
hyperplanes, qneeded approximately to separate N data
points is the order of q = log
2
(N) [3] provided N < 2nand
the computational complexity of this algorithm in Big - O
asymptotic notation is approximately,
O( (
n.Nlog
2
(N) )+ ( n
3
log
2
(N)
) )
where N is the data points and n is the dimension of space.
1. Consider a set of N train points in an n-dimensional
space. Assume this as 'G' space. Consider another
n-dimensional space as 'S', with no points in it. Draw some
initial planes in 'S' space, where the plane equation is
represented as
1 + 1 X 1 + 2 X 2 + 3 X 3 + ... + n X n = 0 (1)α α α α
2. To draw the initial planes, collect the random data
point pairs of size dimension (n) and substitute them in the
above equation (1).
3. Now, calculate the orientation vector (OV) for each point
with respect to all the planes in 'S' space and then transfer
one point at a time from 'G' space to 'S' space. This
Orientation vector gives information about whether the
point lies on the positive or negative side of the plane.
2
Figure 1: Process Flow Diagram
4. While transferring from 'G' space, OV is calculated and
placed in 'S' space for the first point. Then from the second
point onwards, OV is compared with OV of existing points
in 'S' space and if OV of the two points differs then the point
is placed. Which means the point is separated by plane. If
OV of two points matches then point is said to be in
the same ‘quadrant’ [1] and the point is not placed. These
pairs of points are called neighbors and are collected
separately. This pair collection is repeated until it reaches
the size of dimension (n). Please refer to Figure 2 below for
a 2D representation of ‘S’ space.
5. Now, a new plane is drawn in 'S' space by passing
through the midpoints of each and every neighbor pair that
was collected above. This new single plane is sufficient to
separate all the neighbors in 'S' space.
6. Also, whenever a new plane is added, orientation vectors
(OV) of all the existing points in 'S' space are also updated
with respect to new hyperplane generated and this process is
continued until each and every training point is covered.
This ends the process of training the classifier.
7. Now, during predicting the label of a new test point, we
compute the dot product of the test point orientation vector
with all the train points orientation vectors. Take x% of the
maximum value of the dot product results and calculate the
Euclidean distance for those train points with the given test
point and assign the label of the training point with
minimum distance to the test point.
Figure 2: Representation of ‘S’ space point separation in 2D
This is the only algorithm [1] which has the advantage
of incremental learning. That is, if the new points are given
for training, the algorithm starts separating these points from
where it has previously stopped and OV gets updated
accordingly. There is no need to retrain the whole model.
B. Dataset
The model uses two datasets [3] [4], which has been
combined to classify 3 classes as a whole. A total of 108
chest X-ray images of COVID-19 have been taken from this
dataset [3]. The other two types of chest X-ray images
pneumonia and normal are taken from the Kaggle dataset
[4]. Table I shows the distribution of the dataset across the
classes. Though the number of COVID-19 images available
is too small, it had no effect on the model as a whole as
transfer learning-based feature extraction is implemented
and SNN [1] mathematically could separate each data point.
Figure 3 shows the sample image of the dataset.
A B C
Figure 3: Sample chest X-ray images from the datasets. A - COVID-19; B -
Normal; C - Pneumonia
Description
COVID-19
Normal
Pneumonia
Train Set
87
1,270
3,414
Test Set
21
313
859
Table 1: Distribution of Dataset
C. Method
We have used the 4,771 train images to train the KE
Sieve algorithm (as explained in the III-A Architecture
) and
tested on the 1,193 test images. But, as the classifier expects
the inputs to be a real-valued vector, we have done few
pre-processing techniques like converting the image into the
standard size of 224 X 224 and have been fed into the
feature extractor of pre-trained VGG-19 [30] based CNN
model. This represents the feature vector that the KE Sieve
algorithm can consume and separate the train points
distinctly. Principal Component Analysis is applied to the
feature vector and is plotted in 2 dimensions. Figure 4 below
showcases the 2D plot of the feature vector. When this
3
model is tested on the test points it resulted in a new state of
the art accuracy of 98.07%. Fig 1 explains the methodology
used.
IV. EXPERIMENTS
In this paper, the input is the image, and the output is the
predicted class label. The classification algorithm [1] [2] is
implemented in the Python 3 programming language. The
environment is the local system with a processor of Intel®
Core™ i5-8250U CPU @ 1.60GHz × 8 and a RAM of 8 GB
running on Ubuntu OS with no use of GPU.
A. Evaluation Metrics
In this paper, the primary metric is accuracy for model
evaluation. Other metrics like the sensitivity and specificity
are also evaluated for the model. The metrics are measured
using the confusion matrix (Fig 2). Metrics are defined as,
Accuracy = (TP + TN)
/
(TP + FP + FN + TN)
Sensitivity = TP / (TP+FN)
Specificity = TN / (TN+FP)
F
1
- Score = (2 * TP) / ((2 *TP) + FP + FN)
Where, T - True, F - False, P - Positive, N - Negative.
B. Results
The whole model was rebuilt 10 times, as this helps to
keep a check on random initializations.
Fig 3: Confusion Matrix
With the KE Sieve classification approach [1] we just
need 18 planes to separate all the train data points in just ~2
seconds. This resulted in an accuracy of 98.07% and
sensitivity of 100% and specificity of 100% which is greater
than other current models. The other metrics of the model
are shown in Table II. From the confusion matrix it is
observed that there has been misclassification between
Normal and Pneumonia classes. We hope the balanced
dataset may overcome this scenario.
Description
Precision
Recall
F1 - Score
COVID-19
1.00
1.00
1.00
Normal
0.96
0.96
0.96
Pneumonia
0.99
0.99
0.99
Table -II: Metrics Report
The model is built on limited data of COVID-19 is
because the disease is new and datasets are not yet available.
But we are convinced that with more data also, these results
will hold and surely a large number of images would
contribute to improving the model, because of the nature of
the SNN algorithm [1]. From our experience with KE SNN
[1], the more the data points the better the model is as more
planes are drawn to separate the data points.
Figure 4: 2D Representation of Feature Vectors
C. Comparative Study
Our proposed model shows that the classification
performance on the dataset [3] [4] is improved and this
resulted in higher accuracy when compared with the results
of others mentioned in Section II - Related Work.
V. CONCLUSION
The contribution of this paper is showcasing the
effectiveness of this end to end classification model and
chest X-ray images are the best tool for the detection of
COVID-19 and the proposed model classifies between
COVID-19, Healthy, and Pneumonia patients. This model
has been deployed as an application [31] and anyone can
test it by uploading the X-ray image of the chest.
In our future work, we will explore better ways in
feature extraction and classification and also try for the
balanced dataset, to ensure the model is not over-fitted or
tuned to a specific dataset.
4
REFERENCES
[1] K Eswaran. "A non iterative method of separation of points by planes
in n dimensions and its application”. arXiv preprint
arXiv:1509.08742v5, 2015.
[1b] K Eswaran, “An Universal Classifier for Learning and Classification
of data with uses in Machine Learning”. Filed International Patent
with International Patent Geneva, Switzerland. Publication Number -
WO 2017/191648 A1 and Published on 09 November 2017. [29]
[2] K Eswaran. “On non-iterative training of a neural classifier Part-I:
Separation of points by planes.” 10.1109/IntelliSys.2017.8324238,
2017.
[3] Joseph Paul Cohen and Paul Morrison and Lan Dao, COVID-19
image data collection, arXiv preprint arXiv:2003.11597,2020
https://github.com/ieee8023/covid-chestxray-dataset
[4] Chest X-Ray Images (Pneumonia) dataset in Kaggle -
https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia
[5] Roosa, K.,Lee, Y.,Luo, R.,Kirpich,A., Rothenberg, R., Hyman,J.
M., Yan,P., and Chowell, G. Real-time forecasts of the
COVID-19 epidemic in China from February 5th to February
24th, 2020. Infectious Disease Modelling,5: 256-263, 2020.
[6] Yan, L., Zhang, H. T., Xiao, Y., Wang, M., Guo, Y., Sun, C.,
Tang, X., Jing, L., Li, S., Zhang, M., Xiao, Y., Cao, H., Chen, Y.,
Ren, T., Jin, J., Wang, F., Xiao, Y., Huang, S., Tan, X., Huang, N.,
Jiao, B., Zhang, Y., Luo, A., Cao, Z., Xu, H., and Yuan, Y.
Prediction of criticality in patients with severe COVID-19
infection using three clinical features: a machine learning-based
prognostic model with clinical data in Wuhan. medRxiv
preprint doi: https://doi.org/10.1101/2020.02.27.20028027, 1-18,
2020.
[7] Stoecklin, S. B., Rolland, P., Silue, Y., Mailles, A., Campese, C.,
Simondon, A., Mechain, M., Meurice, L., Nguyen, M., Bassi C.,
Yamani, E., Behillil, S., Ismael,S.,Nguyen, D., Malvy, D., Lescure, F.
X., Georges, S., Lazarus, C., Tabaï, A., Stempfelet, M., Enouf, V.,
Coignard, B., Levy-Bruhl, D. and Team, I. First cases of coronavirus
disease 2019 (COVID-19) in France: surveillance, investigations
and control measures, January 2020. Eurosurveillance, 25(6)
:2000094, 2020.
[8] Huang, C., Wang, Y., Li, X., Ren, L., Zhao, J., Hu, Y., Zhang, L.,
Fan, G., Xu, J., Gu, X., Cheng, Z., Yu, T., Xia, J., Wei, Y., Wu, W.,
Xie, X., Yin, W., Li, H., Liu, M., Xiao, Y., Gao, H., Guo, L., Xie, J.,
Wang, G., Jiang, R., Gao, Z., Jin Q., Wang, J., and Cao, B. Clinical
features of patients infected with 2019 novel coronavirus in Wuhan,
China. The Lancet, 395(10223): 497-506, 2020.
[9] https://www.who.int/health-topics/coronavirus
[10] https://www.nationalgeographic.com/science/2020/02/here-is-what-c
oronavirus-does-to-the-body
[11] https://www.who.int/westernpacific/emergencies/covid-19
[12] https://covid19.who.int/
[13] Mahase, E. Coronavirus: COVID-19 has killed more people than
SARS and MERS combined, despite lower case fatality rate. The
BMJ, 368:m641, doi: 10.1136/bmj.m641, 2020.
[14] Wang, Y., Hu, M., Li, Q., Zhang, X. P., Zhai, G., and Yao, N.
Abnormal respiratory patterns classifier may contribute to
large-scale screening of people infected with COVID-19 in an
accurate and unobtrusive manner. arXiv preprint arXiv : 2002.05534,
1-6, 2020.
[15] Ai, T., Yang, Z., Hou, H., Zhan, C., Chen, C., Lv, W., Tao,
X., Sun, Z., and Xia, L. Correlation of Chest CT and RT-PCR
Testing in Coronavirus Disease 2019 (COVID-19) in China: A
Report of 1014 Cases. Radiology, 200642: 1-23,
https://doi.org/10.1148/radiol.2020200642, 2020.
[16] Zhang, W. Imaging changes of severe COVID-19 pneumonia in
advanced stage. Intensive care medicine, 1-3,
https://doi.org/10.1007/s00134-020-05990-y, 2020.
[17] https://www.india.gov.in/india-glance/profile
[18] https://icmr.nic.in/sites/default/files/whats_new/ICMR_testing_updat
e_19April_9PM_IST.pdf
[19] https://ourworldindata.org/covid-testing
[20] Yicheng Fang, Huangqi Zhang, Jicheng Xie, Minjie Lin, Lingjun
Ying, Peipei Pang, and Wenbin Ji. Sensitivity of chest CT for
COVID-19: comparison to RT-PCR. Radiology, page 200432, 2020.
[21] Feng Shi, Liming Xia, Fei Shan, Dijia Wu, Ying Wei, HuanYuan,
Huiting Jiang, Yaozong Gao, He Sui, and DinggangShen.
Large-scale screening of COVID-19 from the community acquired
pneumonia using infection size-aware classification, 2020.
[22] Ophir Gozes, Maayan Frid-Adar, Hayit Greenspan, Patrick D
Browning, Huangqi Zhang, Wenbin Ji,Adam Bernheim, and Eliot
Siegel. Rapid AI development cycle for the coronavirus
(COVID-19) pandemic: Initial results for automated detection &
patient monitoring using deep learn-ing CT image analysis. arXiv
preprint arXiv:2003.05037, 2020
.
[23] Sethy, P. K., and Behera, S. K. Detection of coronavirus Disease
(COVID-19) based on Deep Features. preprints doi:
10.20944/preprints202003.0300.v1, 2020.
[24] Shan, F., Gao, Y., Wang, J., Shi, W., Shi, N., Han, M., Xue, Z., and
Shi, Y. Lung Infection Quantification of COVID-19 in CT
Images with Deep Learning. arXiv preprint arXiv:2003.04655,
1-19, 2020.
[25] Xu, X., Jiang, X., Ma, C., Du, P., Li, X., Lv, S., Yu, L., Chen, Y., Su,
J., Lang, G., Li, Y., Zhao, H., Xu, K., Ruan, L., and Wu, W. Deep
Learning System to Screen Coronavirus Disease 2019 Pneumonia.
arXiv preprint arXiv:2002.09334, 1-29, 2020.
[26] Wang, S., Kang, B., Ma, J., Zeng, X., Xiao, M., Guo, J.,Cai, M.,
Yang, J., Li, Y., Meng, X., and Xu, B. A deep learning algorithm
using CT images to screen for CoronaVirus Disease (COVID-19).
medRxiv preprint doi: https://doi.org/10.1101/2020.02.14.20023028,
1-26, 2020.
[27] Thomas Cherian, E Kim Mulholland, John B Carlin, Harald
Ostenson, Ruhul Amin, Margaret de Campo, David Green-berg,
Rosanna Lagos, Marilla Lucero, Shabir A Madhi, et al.Standardized
interpretation of pediatric chest radiographs for the diagnosis of
pneumonia in epidemiological studies.Bulletin of the World Health
Organization, 83:353–359, 2005.
[28] Kalnad, Nayan, Niranjan Bose, Hari Menon. India’s great healthcare
challenge, also an opportunity. Hindustan Times. June 2017.
http://www.hindustantimes.com/india-news/india-s-great-healthcare-c
hallenge-and-opportunity/story-eB691Pj889x57TYo8MobaM.html
[29] https://patentimages.storage.googleapis.com/be/5a/ae/c34c5da501db2
2/WO2017191648A1.pdf
[30] K. Simonyan, A. Zisserman. “Very Deep Convolutional Networks for
Large-Scale Image Recognition” arXiv preprint arXiv:
1409.1556,
2014.
[31] https://api.alpes.ai/covidprediction/
5
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Background Chest CT is used for diagnosis of 2019 novel coronavirus disease (COVID-19), as an important complement to the reverse-transcription polymerase chain reaction (RT-PCR) tests. Purpose To investigate the diagnostic value and consistency of chest CT as compared with comparison to RT-PCR assay in COVID-19. Methods From January 6 to February 6, 2020, 1014 patients in Wuhan, China who underwent both chest CT and RT-PCR tests were included. With RT-PCR as reference standard, the performance of chest CT in diagnosing COVID-19 was assessed. Besides, for patients with multiple RT-PCR assays, the dynamic conversion of RT-PCR results (negative to positive, positive to negative, respectively) was analyzed as compared with serial chest CT scans for those with time-interval of 4 days or more. Results Of 1014 patients, 59% (601/1014) had positive RT-PCR results, and 88% (888/1014) had positive chest CT scans. The sensitivity of chest CT in suggesting COVID-19 was 97% (95%CI, 95-98%, 580/601 patients) based on positive RT-PCR results. In patients with negative RT-PCR results, 75% (308/413) had positive chest CT findings; of 308, 48% were considered as highly likely cases, with 33% as probable cases. By analysis of serial RT-PCR assays and CT scans, the mean interval time between the initial negative to positive RT-PCR results was 5.1 ± 1.5 days; the initial positive to subsequent negative RT-PCR result was 6.9 ± 2.3 days). 60% to 93% of cases had initial positive CT consistent with COVID-19 prior (or parallel) to the initial positive RT-PCR results. 42% (24/57) cases showed improvement in follow-up chest CT scans before the RT-PCR results turning negative. Conclusion Chest CT has a high sensitivity for diagnosis of COVID-19. Chest CT may be considered as a primary tool for the current COVID-19 detection in epidemic areas.
Article
Full-text available
The initial cluster of severe pneumonia cases that triggered the COVID-19 epidemic was identified in Wuhan, China in December 2019. While early cases of the disease were linked to a wet market, human-to-human transmission has driven the rapid spread of the virus throughout China. The Chinese government has implemented containment strategies of city-wide lockdowns, screening at airports and train stations, and isolation of suspected patients; however, the cumulative case count keeps growing every day. The ongoing outbreak presents a challenge for modelers, as limited data are available on the early growth trajectory, and the epidemiological characteristics of the novel coronavirus are yet to be fully elucidated. We use phenomenological models that have been validated during previous outbreaks to generate and assess short-term forecasts of the cumulative number of confirmed reported cases in Hubei province, the epicenter of the epidemic, and for the overall trajectory in China, excluding the province of Hubei. We collect daily reported cumulative confirmed cases for the 2019-nCoV outbreak for each Chinese province from the National Health Commission of China. Here, we provide 5, 10, and 15 day forecasts for five consecutive days, February 5th through February 9th, with quantified uncertainty based on a generalized logistic growth model, the Richards growth model, and a sub-epidemic wave model. Our most recent forecasts reported here, based on data up until February 9, 2020, largely agree across the three models presented and suggest an average range of 7409e7496 additional confirmed cases in Hubei and 1128e1929 additional cases in other provinces within the next five days. Models also predict an average total cumulative case count between 37,415 and 38,028 in Hubei and 11,588e13,499 in other provinces by February 24, 2020. Mean estimates and uncertainty bounds for both Hubei and other provinces have remained relatively stable in the last three reporting dates (February 7th e 9th). We also observe that each of the models predicts that the epidemic has reached saturation in both Hubei and other provinces. Our findings suggest that the containment strategies implemented in China are successfully reducing transmission and that the epidemic growth has slowed in recent days.
Article
Full-text available
A novel coronavirus (severe acute respiratory syndrome coronavirus 2, SARS-CoV-2) causing a cluster of respiratory infections (coronavirus disease 2019, COVID-19) in Wuhan, China, was identified on 7 January 2020. The epidemic quickly disseminated from Wuhan and as at 12 February 2020, 45,179 cases have been confirmed in 25 countries, including 1,116 deaths. Strengthened surveillance was implemented in France on 10 January 2020 in order to identify imported cases early and prevent secondary transmission. Three categories of risk exposure and follow-up procedure were defined for contacts. Three cases of COVID-19 were confirmed on 24 January, the first cases in Europe. Contact tracing was immediately initiated. Five contacts were evaluated as at low risk of exposure and 18 at moderate/high risk. As at 12 February 2020, two cases have been discharged and the third one remains symptomatic with a persistent cough, and no secondary transmission has been identified. Effective collaboration between all parties involved in the surveillance and response to emerging threats is required to detect imported cases early and to implement adequate control measures.
Patent
Full-text available
The present invention discloses a new methodology which can be used to classify specific problems and describes a product such as the “classification engine” which is implementable in hardware for specific problems. The specific problems could be such as Face Recognition Systems, Disease diagnostic systems, Robotic Inspection systems for use in the factory shop floor etc. The system and method given in this invention examines any form of data for the purposes of learning and classification so that any Machine such as a robot or machine will be able to handle the data and classify the data for automatic decision making. While analyzing the data it performs, basically THREE main steps (I) The Partition Process and (II) The Reduction of Dimension (of data) Process, (III) The Cluster Discovery Process and (IV) The Classification Process. See pages 27 to 33 for the written opinion of the International Authority.
Conference Paper
Full-text available
In this paper and the next an entirely novel method of supervised neural learning which is non-iterative is described. The process is as follows: every data point which may be n-dimensional is first separated from every other point by hyper-planes, so that no two points are un-separated by at least one hyper plane. This is always possible in high dimension spaces because of the availability of many degrees of freedom. The separation is done by a newly discovered non-iterative algorithm and is described in this paper. Once the data points are all separated by planes, these planes are then used to classify the data points. If sufficient information is known as to which point belongs to which class then, one can regroup (or cluster) those points which belong to the same class. Since the equations to the planes which separate each point from each other, are now known, one can use some of these planes to separate each cluster (if needed one may have to add a few more planes). So that one has a set of planes which separate each cluster from one another. This paper describes a set of algorithms which perform this clustering and which discover the planes separating each cluster from another, these algorithms once again perform the task in a non-iterative manner. The planes that separate each cluster can then form the basis of a neural architecture. Thus in this manner one obtains a Neural architecture to solve the classification problem. This paper and the next describes the entire process. However, this first paper confines itself to the separation of points and describes the Separation Algorithm which can separate any number of points in n-dimensional space from one another by hyper planes. Given a set G of N f points, along with their coordinates, in n-dimensional X-Space, the algorithm partitions all the N f points by using planes such that no two points are left un-separated by some plane. And and the next paper deals with the clustering problem and the determination of a neural architecture. Example problems are solved which demonstrate the efficiency and practicability of this entirely new method of classification .
Article
Full-text available
Although radiological pneumonia is used as an outcome measure in epidemiological studies, there is considerable variability in the interpretation of chest radiographs. A standardized method for identifying radiological pneumonia would facilitate comparison of the results of vaccine trials and epidemiological studies of pneumonia. A WHO working group developed definitions for radiological pneumonia. Inter-observer variability in categorizing a set of 222 chest radiographic images was measured by comparing the readings made by 20 radiologists and clinicians with a reference reading. Intra-observer variability was measured by comparing the initial readings of a randomly chosen subset of 100 radiographs with repeat readings made 8-30 days later. Of the 222 images, 208 were considered interpretable. The reference reading categorized 43% of these images as showing alveolar consolidation or pleural effusion (primary end-point pneumonia); the proportion thus categorized by each of the 20 readers ranged from 8% to 61%. Using the reference reading as the gold standard, 14 of the 20 readers had sensitivity and specificity of > 0.70 in identifying primary end-point pneumonia; 13 out of 20 readers had a kappa index of > 0.6 compared with the reference reading. For the 92 radiographs deemed to be interpretable among the 100 images used for intra-observer variability, 19 out of 20 readers had a kappa index of > 0.6. Using standardized definitions and training, it is possible to achieve agreement in identifying radiological pneumonia, thus facilitating the comparison of results of epidemiological studies that use radiological pneumonia as an outcome.
Preprint
Background The outbreak of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-COV-2) has caused more than 2.5 million cases of Corona Virus Disease (COVID-19) in the world so far, with that number continuing to grow. To control the spread of the disease, screening large numbers of suspected cases for appropriate quarantine and treatment is a priority. Pathogenic laboratory testing is the gold standard but is time-consuming with significant false negative results. Therefore, alternative diagnostic methods are urgently needed to combat the disease. Based on COVID-19 radiographical changes in CT images, we hypothesized that Artificial Intelligence’s deep learning methods might be able to extract COVID-19’s specific graphical features and provide a clinical diagnosis ahead of the pathogenic test, thus saving critical time for disease control. Methods and Findings We collected 1,065 CT images of pathogen-confirmed COVID-19 cases (325 images) along with those previously diagnosed with typical viral pneumonia (740 images). We modified the Inception transfer-learning model to establish the algorithm, followed by internal and external validation. The internal validation achieved a total accuracy of 89.5% with specificity of 0.88 and sensitivity of 0.87. The external testing dataset showed a total accuracy of 79.3% with specificity of 0.83 and sensitivity of 0.67. In addition, in 54 COVID-19 images that first two nucleic acid test results were negative, 46 were predicted as COVID-19 positive by the algorithm, with the accuracy of 85.2%. Conclusion These results demonstrate the proof-of-principle for using artificial intelligence to extract radiological features for timely and accurate COVID-19 diagnosis. Author summary To control the spread of the COVID-19, screening large numbers of suspected cases for appropriate quarantine and treatment measures is a priority. Pathogenic laboratory testing is the gold standard but is time-consuming with significant false negative results. Therefore, alternative diagnostic methods are urgently needed to combat the disease. We hypothesized that Artificial Intelligence’s deep learning methods might be able to extract COVID-19’s specific graphical features and provide a clinical diagnosis ahead of the pathogenic test, thus saving critical time. We collected 1,065 CT images of pathogen-confirmed COVID-19 cases along with those previously diagnosed with typical viral pneumonia. We modified the Inception transfer-learning model to establish the algorithm. The internal validation achieved a total accuracy of 89.5% with specificity of 0.88 and sensitivity of 0.87. The external testing dataset showed a total accuracy of 79.3% with specificity of 0.83 and sensitivity of 0.67. In addition, in 54 COVID-19 images that first two nucleic acid test results were negative, 46 were predicted as COVID-19 positive by the algorithm, with the accuracy of 85.2%. Our study represents the first study to apply artificial intelligence to CT images for effectively screening for COVID-19.