DEEP LEARNING: FIGHTING COVID-19 WITH NEURAL NETWORKS

OVERVIEW

Deep learning is a part of AI which was founded on artificial neural organizations, as neural organization will reflect the human thoughts. Deep learning which a subset of Machine Learning is is making revolutions in each field it comes into play. It decreases the human work and accomplishes the work precisely than people do. It emulates the human Brain with artificial neural organizations and dominating in all the fields where it is deployed. In this article, we primarily center on the deployment of deep learning in these hard COVID times, overview and the breakthroughs of deep learning in all aspects of it.

1. INTRODUCTION
   1. Deep Learning

Deep Learning involves solving the complex decisions without human interference and giving the finest output out of the unstructured input data. In Deep learning, we don't have to explicitly program everything. A neural network is a “progression of calculations that tries to perceive basic relationships in a lot of information through a process that impersonates the manner in which the human cerebrum works”. ’Deep’ implies the compound layers of network.

1.2 Architectures

1. Recurrent Neural Network

Recurrently uses the output of previous layers and proceeds further by giving the same as input for the next layer. It was mainly used in the unsegment jobs such as Speech Recognition and Handwriting Recognition.

2. Convolution Neural Network

           It is most commonly employed in Visual images by classifying, finding the features and various aspects of the same. It finds the relationships between the pixels in the image by mathematical calculations. It outperforms by applying various filters and finds the essential components vital for the computations. It is applied in image recognition, Object detections and facial recognitions.

3. Deep Belief Network (DBN)

DBN has the greater capability to recognize and distinguish patterns. It consists of Restricted Boltzmann Machine (RBM) stacked together in the middle layer with classifier as the final layer. It uses probabilities and unsupervised learning to produce the precise output. Each layers absorbs the whole input and start processing. It is used in image recognition, Video sequence recognition.

2. DEEP REINFORCEMENT LEARNING

Reinforcement Learning is a process of learning through trial and error method, with some rewards and punishments. The agents learn and construct their own rules directly from the inputs in the environment. For Example: In the chess game, Agent is one of the players; Environment includes the board and the other player.

Deep learning helps in analyzing complex input and giving the best response. The "deep" segment of reinforcement learning raises to different (deep) layers of neural organizations that imitate the structure of a human mind. Deep learning requires a lot of preparing data and huge supposing power. In the course of the most recent couple of years, the volumes of information have exploded while the expenses for figuring power have significantly decreased, which has empowered the blast of deep learning applications. Deep reinforcement learning is the assortment of deep learning and reinforcement learning. This type is used in making very complex decisions. Majorly it is applied in the areas like Education, Robotics, NLP, Transportation etc.

3. SOME WELL- KNOWN APPLICATIONS

Some of the incredible applications of deep learning are “NLP, speech recognition, face recognition, Object detection etc., For instance, while you add a photograph together along with your friend on Facebook, Facebook routinely tags your friends and shows you his name”. Facebook makes use of deep learning strategies to understand a face.

• Deep learning in healthcare allows in discovery of drugs and their development.
• Medical imaging strategies along with MRI scans, CT scans, ECG, are used to diagnose dreadful illnesses along with coronary heart sickness, cancer, brain tumor.  
• Deep learning is used to investigate the medical health insurance fraud claims.
• Deep learning strategies recognize human spoken languages and convert them into text.

4. BREAKTHROUGHS BY DEEP LEARNING IN COVID-19

Deep learning is helping clinical specialists and analysts to discover the covered potential outcomes in data and to serve the medical care venture better. Deep learning in healthcare affords doctors, the evaluation of any sickness appropriately and allows them deal with them higher, as a result ensuing in higher clinical decisions.

Covid disease (COVID-19) is an “irresistible illness brought about by a newfound Corona Virus”. The vast majority who fall hampered with COVID-19 will encounter mild to moderate indications and recover without special treatment. Dominant part of the public’s is not open to the effect of this ailment and very little expectation on their well-being and defenseless to the disease without any problem. Deep Learning Algorithms can be employed and thus lessen the threat to the public by reducing the human connect and systematize every solitary progression.

4.1 Detect Covid-19 from Sound of Cough

Artificial intelligence-based system can be built which will hear your cough and indicate whether you have got COVID-19. Such model can be developed by collecting cough sounds from COVID-19 patients as well as from those persons who don't seem to be infected by the virus. An enquiry using AI is conducted on the mixed information of cough sound samples to find out a corona virus infected person that reciprocally will find the person with additional probabilities of obtaining the infection. This will assist the public’s to analyze themselves and they can make a move forward if they are getting high chances of risk in that prediction.

4.2 Cough Recognition in Public Places

The coughing detection camera can be installed in public places which recognizes wherever coughing happens, envisioning the locations and the picture of the person. The cough recognition camera will track the info regarding the one who coughed, their location, and also the variety of coughs on a period of time basis. The coughing can be taught to the model by uploading coughing sample images and can easily differentiate between coughing or not through the video sequence recognition using Deep Belief Network. This new technology is estimated to be very useful for police investigation that can inhibit epidemic transmissions in an exceedingly non-contact approach. The model visualizes the cough event and specifies the locality and some more information very clearly.

4.3 Predict Covid Through X-Rays

The drawbacks of manual testing include thin accessibility of testing kits, expensive and inefficient check; a blood test takes around 6–7 hours to come up with the result. So, the concept is to beat these circumstances using the DL technique for higher and economical treatment. Since the sickness is extremely contagious so as early as we tend to generate the results, the less cases within the city. CNN can accurately predict the COVID than human do. This system can be employed in unsociable places in countries put down low with COVID-19 to beat unavailability of radiologists. Such models are also used to diagnose other chest-related illnesses together with Tuberculosis and respiratory disorder.

4.4 Prediction Without Testing

Manual testing is done to distinguish the COVID presence in human body, yet it is a tedious process to check for every individual physically and it might lead to the spread of this irresistible infection. Instead We can develop a model that strongly predict the persons with more symptoms of COVID-19.The classification is mainly done on the basis of the most common symptom of the disease. We can utilize the symptoms of COVID and the test results of COVID patients as information for the model which will unquestionably limit the zone to be engaged and the individual to be focused. On examining which symptoms well on the way to be related with a positive test, for sure loss of taste and smell (anosmia) was especially striking, with 66% of users testing positive for COVID disease revealing this side effect contrasted and a little more than persons who tested negative. The discoveries recommend that anosmia is a more grounded indicator of COVID-19 than fever. A model can be built up that can predict whether an individual is probably going to have COVID-19 dependent on their age, sex and a blend of four key manifestations: loss of smell or taste, serious or tenacious cough, weakness and skipping suppers. We can easily find the patterns in the symptoms and thereby change the most common symptom based on the predictions or we can add up the symptom with the existing one. It paves way for us to easily recognize persons with more danger and the persons who are in need of immediate treatment based on the symptoms they have.

5. TACKLE COVID-19 USING MACHINE LEARNING

Machine Learning technologies and techniques are used for enhancing the prediction accuracy. Here, the screening is done based on infectious and non infectious disease. The connection between medical care shows development of the initial expert system known as MYCIN (1976). This is nothing but an antibiotic for patients to make a treatment for bacterial infection. This expert system helps in handling clinical decisions for experts in the medical field. In this study, the Machine learning technologies are used for numerous epidemic outbreaks. This contributes a huge advantage on medical experts to identify the high level communicable diseases such as SARA, HIV, EBOLA and COVID-19. This also shows non-communicable diseases like Cancer, Heart disease, Diabetes and Stroke more which clearly leads to an outbreak [15] – [20].

5.1 Machine Learning Technology for Screening Covid-2019

In general, diseases are classified into two categories which are infectious and non infectious. Rapid screening assists in preventing the spread of Coronavirus disease with effective cost and speed up diagnosis which is the most important task to save more human lives. An expert system is developed by medical-care that helps in identification of screening management for COVID-19 outbreak contrast with the traditional method. Machine learning is used for increasing the recognition and screening methods to analyze the diseased patient with specific tools like Radio Imaging, Computed Tomography which is CT image, X-Rays and data of blood samples. The accomplishment of the device holds up a moderate position in the outburst of Covid-2019 pandemic. By considering the ML tools a new model is validated to construct a method on Deep Convolutional Network for the screening test of Coronavirus-2019. This takes advantage of an expert system in employing Machine learning CT images to obtain the Covid-19 infected peoples to perform with radiological tools that result in 86.27% accuracy respectively. In recent studies, an auxiliary tool is used for developing the best accuracy of Coronavirus which is identified and detected by machine learning algorithms [21] – [26].

The constructed model assists in X-ray image of infected patients’ chest in order to find the pneumonia cases and store the records. Here, the performance of the accuracy is incredible for validating the process of screening with accurate radiology operation (shown in the figure 3). In addition, researchers recognized four key combinations such as Clinical feature, Laboratory feature, demographic data and support vector machine for primary classification of models. The new model helps in predicting the critical condition of patients that result with training and testing of datasets. In screening, the healthcare experts are more benefited with auxiliary tools. The quick detection assists in decreasing the spread of the disease and engages more time on medical experts to find out the next diagnosis to save the lives of the infected patients with minimal cost. This article focuses majorly on classification of algorithms or single information which comes up with hybrid operation methods with specific algorithms on databases. At the end, this shows the true identity of affected patients and categorization of applications to the real world.   

5.2  Contact Tracing of Covid-19

If  Covid-19 the particular disease is confirmed for the patient then the next necessary step is contact tracing. This prevents the spread of disease on a large scale. The World Health Organization has stated that the spread of Coronavirus is through the interconnections between person to person. These are primarily affected from nose saliva, droplets and discharges in contact with transmission. In order to manage, contact tracing is more important in health-care tools for breaking out the chain transmission of the virus. Contact tracing is the process of identifying and managing infected peoples over Covid-19 in occurrence to control the exposure of the epidemic spread. This suppresses the transmission of novel coronavirus to outbreak and decrease the magnitude of the pandemic. Numerous countries are coming out with digital contact tracing which helps in processing through digitization modes such as mobile apps, bluetooths, global positioning system,  social graph, application program interface, mobile tracking, data transaction, social networks, card transactions, system address and physical operations of systems.

This performs as real-time virtual reality with much faster processing speed compared with other non digital systems. These digital apps are developed in order to collect and manage personal data of infectious patients. Further, it is determined by Machine learning tools that trace out the infected person across the connected chain of novel coronavirus-19. The figure (iv) shows up the two parameters such as contact and test. The contact holds up the connection between two persons named as person A and person B that exchanges a key code through bluetooth automatically when they come closer similarly in the test phase, any one of the infected people such as person A initiated to have the symptoms are advised to take up Covid test. The next step includes Decentralized and centralized approaches. Here, the patient needs to feed their data with location and a Machine learning algorithm analyses the interaction of data and process who should be contacted to send out the alerts. Finally, the Covid positive infected persons are encountered to proceed with treatment.

5.3 Predicting and Forecasting of Covid-19

Machine Learning helps in forecasting and predicting selective information for novel epidemics. In recent work, supervised multi-layered recursive classifiers are used for analysing the clinical dataset factors. This processes an expert alert system for medical purposes to tackle the cause of pandemic. The decision rules are focused on a short term memory network for developing the time series of the forecasting model. Here, the Covid-19 collects the data and information from highly infected countries. Then, the transformation of data is trained and curved using a Machine learning algorithm called ANN which is Artificial Neural Network. The disease is predicted based on the transmission of infected patients and the data are forecasted based on the development of a robust forecast model for controlling COVID-2019 cases (shown in the figure 5).

5.4 Drug Vaccines for Coronavirus

The researchers and healthcare experts initiated to develop and tackle drug vaccination for coronavirus in order to decrease the outbreak of the infectious disease. Machine learning technology concerns the drug possibility of choice for the infected patient's treatment by managing the existing medicines for the essentials of human beings. This article acquires numerous technologies and methodologies labeling the classification of advanced ML algorithms based on statistical data. This assists tool computation and docking applications to predict reusable or existing drugs for Covid-19 medications. The development of drugs is the risk factor and cost effective process. The discovery of the treatment and medication draws security with clinical data. The challenges are comprehensively selected for hybrid data and develop a real life application for vaccination and drug management.

6. CONCLUSION

Deep Learning is composed of artificial neural networks for solving complex problems and make quicker decisions easily. It has wide range of applications in all the departments. Coming to the healthcare, contributions of deep learning is numerous and it is still continuing more. Even during these rigid COVID times, they paved way for us to self-diagnose using the tools deployed using DL either one through the parameters such as cough, symptoms or X-Rays.

REFERENCES

1. Balaji, B.S., Balakrishnan, S., Venkatachalam, K. et al. Automated query classification based web service similarity technique using machine learning. J Ambient Intell Human Comput (2020). https://doi.org/10.1007/s12652-020-02186-6.
2. Balakrishnan S, K.Aravind, A. Jebaraj Ratnakumar, “A Novel Approach for Tumor Image Set Classification Based On Multi-Manifold Deep Metric Learning”, International Journal of Pure and Applied Mathematics, Vol. 119, No. 10c, 2018, pp. 553-562.
3. S. Balakrishnan, D.Deva, ”Machine Intelligence Challenges in Military Robotic Control”, CSI Communications magazine, Vol. 41, issue 10, January 2018, pp. 35-36.
4. A. Jebaraj Rathnakumar, S.Balakrishnan, “Machine Learning based Grape Leaf Disease Detection”, Jour of Adv Research in Dynamical & Control Systems, Vol. 10, 08-Special Issue, 2018. Pp. 775-780.
5. D.Prabha, R. Siva Subramanian, S.Balakrishnan, M.Karpagam, Performance Evaluation of Naive Bayes Classifier with and without Filter Based Feature Selection, International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-8 Issue-10, August 2019, pp. 2154-2158.
6. S. Vasu, A.K. Puneeth Kumar, T. Sujeeth, Dr.S. Balakrishnan, “A Machine Learning Based Approach for Computer Security”, Jour of Adv Research in Dynamical & Control Systems. Vol.10, 11-Special issue, 2018, pp. 915- 919.
7. M. Balasubramanian, S.Balakrishnan, Reflections on COVID-19 in India: A Survey, International Journal of Pharmaceutical Research, 12 (4), 137-140.
8. D. Winston Paul, S. Balakrishnan, A.Velusamy, Rule Based Hybrid Weighted Fuzzy Classifier For Tumor Data, International Journal of Engineering and Technology (UAE). Vol. 7,  (4.19) (2018), pp. 104-108 Doi: https://doi.org/10.14419/ijet.v7i4.19.
9. Balakrishnan S, M.Syed Shahul Hameed, Shameer Mohammed, Shimaz Khan Shaik, Aswin G, SARS-CoV-2 Trauma Handling and Hazard Rescue Through Machine Learning, International Journal of Pharmaceutical Research. - Mar 2021, Vol. 13, Issue 1, pp. 3302 – 3308. https://doi.org/10.31838/ijpr/2021.13.01.500
10. Daniel Francis Selvaraj, Balakrishnan S, Manopaul P, Coronavirus Disease (Covid-19) Outbreak In India, International Journal of Pharmaceutical Research, Jan - Mar 2021, Vol. 13, Issue 1, pp. 568-571.  https://doi.org/10.31838/ijpr/202 1.13.01.086  
11. R. Prabha, Balakrishnan S, S. Deivanayagi, V.K.G. Kalaiselvi, D. Pushgara rani , Aswin G, A Review of Classification Algorithms in Machine Learning for Medical IoT, International Journal of Pharmaceutical Research. Jan - Mar 2021, Vol. 13, Issue 1, pp. 3000 – 3007. https://doi.org/10.31838/ijpr/2021.13.01.448
12. 20. S. Balakrishnan, Aswin G. Deep Learning: An Overview and its Practical Examples, CSI Communications magazine, Volume No. 44, Issue No. 8, November 2020, pp. 17-19.
13. 21. Balakrishnan S, K.Aravind, A. Jebaraj Ratnakumar, “A Novel Approach for Tumor Image Set Classification Based On Multi-Manifold Deep Metric Learning”, International Journal of Pure and Applied Mathematics, Vol. 119, No. 10c, 2018, pp. 553-562.
14. 22. S. Balakrishnan, Aswin G. Deep Learning: An Overview and its Practical Examples, CSI Communications magazine, Volume No. 44, Issue No. 8, November 2020, pp. 17-19.
15. 23. S. Balakrishnan, J. Janet, Aswin G. Artificial Intelligence and Robotics: A Research Overview, CSI Communications magazine, Volume No. 44, Issue No. 7, October 2020, pp. 24-26.
16. 24. Balakrishnan. S, Artificial Intelligence in a Quantum World of Computation, CSI Communications magazine, Volume No. 44, Issue No. 3, June 2020, pp. 28-30.
17. 25. Balakrishnan. S and K L Shunmuganathan. Article: A JADE Implementation of Integrated Agent System for E-Mail Coordination (IASEC). International Journal of Computer Applications 58(5): 5-9, November 2012.
18. 26. S.Balakrishnan, “An Overview of Agent Based Intelligent Systems and Its Tools”, CSI Communications magazine, Volume No. 42, Issue No. 10, January 2019, pp. 15-17.
19. 27. S.Balakrishnan, “Agent based Bioinformatics Integration using JADE – Overview”, CSI Communications magazine, Volume No. 43, Issue No. 9, December 2019, pp. 29-31.
20. 28. Balakrishnan S and Steven Uatuaromuinjo Tjiraso, “Integration of Agent Based Computing with Cloud Computing: Towards Cloud Intelligent Systems”, International Research Publication House, Delhi. Engineering and Technology: Recent Innovations & Research, ISBN- 978-93-86138-06-4, pp. 1-17.
21. 29. S. Balakrishnan, K.N. Sivabalan and J. Janet “MASFE - Mutliagent System for Filtering E-Mails Using JADE”, Advanced Engineering Research and Applications (AERA), Research India Publications, ISBN- 978-93-84443-42-9, pp. 148-167, 2017.
22. 30. P.Arivazhagan, Balakrishnan. S and K L Shunmuganathan. “An Agent Based Centralized Router with Dynamic Connection Management Scheme Using JADE”, International Journal of Applied   Engineering Research, ISSN 0973-4562, Volume 11, Number 3 (2016) pp 2036-2041.
23. 31. Balakrishnan. S and K L Shunmuganathan, R. Sreenevasan, “Amelioration of Artificial Intelligence using Game Techniques for an Imperfect Information Board Game Geister” International Journal of Applied Engineering Research (IJAER). ISSN 0973-4562. Vol 9, Number 22 (2014) pp. 11849-11860.
24. 32. Balakrishnan. S and K L Shunmuganathan, An Agent Based Collaborative Spam Filtering Assistance Using JADE”, International Journal of Applied   Engineering Research, ISSN 0973-4562, Volume 10, Number 21 (2015) pp 42476-42479.
25. 33. A.Jebaraj Rathnakumar, S.Balakrishnan, Design Of Multi-Agent Based Systems For Entrusted Communication Using JADE”, Taga Journal of Graphic Technology, Vol. 14, pp. 766-774, 2018. 
26. 34. S. Balakrishnan J. Janet, Aswin G, Artificial Intelligence and Robotics: A Research Overview, CSI Communications magazine, Volume No. 44, Issue No. 7, October 2020, pp. 24-26.