Blog – AI in the war against COVID-19
Utility of AI in managing COVID-19
Artificial intelligence and big data have emerged as key players in the battle against COVID-19, upping the ante by providing solutions that have demonstrated a jaw dropping escalation of efforts. Humanity has faced several contagious illnesses throughout history, and starting from the late 20th Century, AI has been leveraged against several outbreaks.
AI is being leveraged in the detection of illness clusters, surveillance of cases, prediction of upcoming outbreaks and further mutations, mortality risk, in diagnosing the illness itself, and in disease management through the allotment of resources, enabling training, documentation upkeep, and pattern recognition for analyzing the trajectory the outbreak is taking. Various deployments of AI are witnessing a lot of interest and are providing hope in a bleak situation.
AI for predictive purposes and tracking
AI can be leveraged for foretelling the spread of the virus and for producing early warning frameworks through the extrapolation of data from social media outlets, calls, and news websites and impart useful data about the susceptible regions and for predicting morbidity and mortality. Bluedot detected a cluster of pneumonia cases and foretold the outbreak and geolocation of the COVID-19 outbreak on the basis of current data with the help of machine learning (ML) algorithms. HealthMap collates the publicly distributed data with regards to COVID-19 and makes it easily accessible to enable practitioners to stay on top of spread rates. Lately, the role of artificial intelligence in detecting and predicting COVID-19 outbreaks by deploying multitudinal and multimodal information has been the focus.
AI in contact tracing
AI can supplement mobile health applications where smart gadgetry such as wearables can be leveraged for diagnosing, contact tracing, and effective monitoring of COVID-19. Apps like AI4COVID-19 that are dependent on audio recording sampling of 2 s cough can be leveraged in telemedicine.
AI to monitor COVID-19 cases
AI strategies are being leveraged for monitoring patients in medical scenarios and for predicting the trajectory of the treatment. On the basis of information obtained from critical stats and clinical parameters, AI may furnish vital data for allotment of resources and decision making by prioritizing the requirement for ventilators and respiratory assistance in ICUs. AI can also be leveraged for forecasting the odds of recovering from the illness or mortality in COVID-19 and to furnish everyday updates, storage, and trend analysis and planning the course of treatment.
AI in preliminary diagnosis
AI was leveraged for the identification and quantification of COVID-19 cases from X-ray and CT scan imagery. Analysts have generated a deep learning framework referred to as the COVID-19 detection neural network (COVNet) for distinguishing between COVID-19 and socially acquired pneumonia on the basis of visual 2D and 3D features extrapolated from volumetric chest CT scans. Singh et al. generated a unique deep learning framework leveraging Multi-objective Differential Evolution and convolutional neural networks for COVID-19 diagnoses with the help of a chest scan. COVID-ResNet which was produced leveraging automatic and discriminative learning rate and progressive picture resizing demonstrated improved performance in comparison to COVID-Net in the diagnosis of COVID-19. Alom et al. produced a framework referred to as COVID_MTNet through the application of enhanced Inception Recurrent Residual Neural Network and NABLA-3 network frameworks for identification and localization of areas of interest from x-ray imagery and chest CT scan imagery. Yet another research leveraged AI-based classifiers for forecasting the results of RT-PCR of COVID-19 cases leveraging 16 parameters obtained from comprehensive blood profiling. This may find utility in minimizing the number of RT-PCR tests in environments that lack adequate resources.
AI minimizes the load from doctors and allied healthcare staff
AI-based triage systems can assist in minimizing the work load on medical professionals and healthcare employees through the automation of various processes like providing training to medical professionals, identification of the mode of treatment and care by analyzing clinical data leveraging pattern detection methodology, digitization of patient documentation and additionally by providing solutions that reduce contact time with patients. AI can be leveraged for classifying patients based on intensity of symptoms, genetic predisposition and clinical reports in various categories like mild, moderate, and severe, so that various strategies and approaches can be taken up for managing the patients in the most streamlined fashion. AI within telemedicine can also be leveraged to eradicate the requirement of constant and unneeded clinical visits by remote monitoring of cases and documentation of patient’s information in asymptomatic cases or patients with reduced degree of symptoms. AI-based medical chatbots can additionally be leveraged for consulting purposes, therefore lessening the overcrowding of healthcare facilities which additionally curbs the spread of the illness thus averting the excess load which cripples effective operation of vital health care services. Chatbots such as Clara produced by the CDC and Zini are giving the much-needed assistance to patients in remote environments through extracted features obtained from the information of other patients as training datasets. A similar strategy was leveraged to forecast the probability of getting acute respiratory distress syndrome. Service robots and anthropomorphic robots with AI Core can be leveraged for delivering critical services and mundane tasks like cleansing, disinfecting, and surveillance in clinical settings.
AI in protein structure prediction
AI can assist in forecasting the structure of critical proteins vital for virus entry and replication and impart important feedback that can create a path for drug production in a reduced time frame. AlphaFold algorithm of Google Deep Mind used deep residual networks (DRN) referred to as ResNets for forecasting protein structures of membrane protein, protein 3a, nsp2, nsp4, nsp6, and papin-like C-terminal domain of SARS-CoV-2, which will provide a huge thrust to drug invention programs. DeepTracer, an application based on custom deep convolutional neural networks, was leveraged to obtain protein complex structure of SARS-CoV-2 from high-res cryoelectron microscopy density maps and amino acid sequences.
AI in production of therapeutics
AI strategies can enhance and supplement conventional tech by minimizing the time needed in getting a drug from bench to bed by increasing the speed of lead discovery, virtual screening, and validation procedures by a large extent. AI can also quicken the pace by deriving critical information for drug repurposing or drug repositioning by screening attributes of medicines that have already undergone approval and validation on the basis of molecular descriptors and attributes, which cannot be carried out by a human subject matter expert. BenevolentAI leveraged ML strategies to quicken its medicine discovery program and detected baricitinib as a possible medicine in the fight against COVID-19. Insilico Medicine has detected several miniscule molecules against COVID-19 leveraging AI. Another research used virtual screening and supervised learning to detect possible medications against COVID-19. Zhou et al. took up an integrative network-based systems pharmacological strategy for identifying possible medicines for SARS-CoV-2 from current reservoir of drug molecules and medicine combos. Various other AI-based efforts including inclProject IDentif.AI (identifying infectious illnesses combination therapy with AI) and PolyPharmDB have been useful in detecting candidates against COVID-19. Various ML strategies and deep-learning based applications are also being leveraged for expediting the drug discovery procedures.
AI in production of vaccines
The pace at which at a vaccine is expected to be developed for COVID-19 is unprecedented. This pace can become quicker manifold by leveraging the capabilities of AI. Ong et al. forecasted potential vaccine candidates for COVID-19 leveraging the Vaxign reverse vaccinology-ML platform that is reliant on supervised classification models.
AI in stopping the spread of misinformation
Owing to the overload of data on the matter, the pandemic has been converted into an infodemic. Comprehension of knowledge, awareness and procedures towards COVID-19 by leveraging data from social media can assist in developing the techniques to assemble and distribute timely and accurate data for reducing the impact of COVID-19. ML strategies can be leveraged to detect trends and sentiment analysis and impart data with regards to the origins false data and assist in curbing the rumors and false data. AI strategies can further be leveraged for depicting a clear picture of recovery rates, accessibility, and availability to healthcare and detection of gaps. AI can impart updated news with regards to the surfacing evidence with regards to diagnosing, treating, spectrum of symptoms, and treatment outcomes in this increasingly dynamic scenario, which will assist practitioners in real-world settings and assist the public in getting over fear and hysteria.
AI within genomics
Randhawa et al. developed a procedure for quick and accurate categorization of available SARS-CoV-2 genomes through application of ML on detected genomic signatures. Wang et al. leveraged ontology-based side effect forecasting framework an ANNs to test the side effects of conventional Chinese medications for the treating of SARS-CoV-2.
Conclusions and perspectives looking ahead
Taking up a strategy based on evaluation, isolation, and contact tracing is needed in the battle against COVID-19. It is required to exploit present know-how to produce effective chemotherapeutic agents against COVID-19, obtaining signs from lessons learnt in history in the course of similar outbreaks.
There is no magic treatment against the illness. The need of the hour is to quicken progression on all regards from surveillance and monitoring to prevention and treatment. As COVID-19 marks the third outbreak of a coronavirus in recent history and several other similar viruses are circulating in animal reservoirs, we must concentrate on decoding the molecular mechanism of SARS-CoV-2 and other such viruses and improving our preparations through capacity development for aversion of other such upcoming outbreaks. As the present scenario dictates the requirement for fast-tracked delivery of solutions, reactions to this outbreak were largely augmented by several electronic tech and AI. AI was discovered to be equivalent to and perhaps even more accurate than human practitioners in diagnosing COVID-19 and medicine identification. The need of the hour is larger datasets to train AI models and a lawful framework and ethical aspects for the sharing of information before AI takes the frontlines in diagnosing and other spheres. Various bottlenecks in leveraging AI to its complete potential in the present scenario are availability and the capacity to share clinical and epidemiological information, computational resources, scalability, privacy, and ethical considerations.