Tracking and Forecasting the COVID-19 Outbreak in Real Time: Role of AI

COVID-19 is a highly transmittable viral infection from human to human and has a relatively long incubation period (Q. Li et al. 2020; Zou et al. 2020). During this long incubation period, infected persons might show mild or no symptoms. This is a major challenge from a disease containment point of view', and the role of a forecasting model becomes hugely important, since it will help in identifying potentially high-risk zones and super-spreaders so that preventive measures can be taken. Indeed, forecasting models for growth and geographic distribution of the outbreak, dynamics of transmission, targeting resources, and evaluating the impact of intervention strategies and the evolution of hot spots in the outbreak are a very important aspect. All of these may be based upon ML and AI, which might turn out to be very useful tools for understanding the epidemiology of the viral infection and predict its potential impact on public health in local and global communities (Abhari et al. 2020; Hu et al. 2020).

These models can help governments or regulatory bodies to understand how' COVID-19 outbreaks evolve, and control measures that need to be taken/altered for controlling the outbreak including source containment, case management, policy directives for schools and organizations, contact tracing, protocols for closing borders, suspending/limiting community services, infection control at healthcare facilities, and community containment (Perrella et al. 2020; Zeng et al. 2020). Globally, serious efforts are going on in the direction of tracking the COVID-19 outbreak with the assistance of Al, and some of them are sampled in the following sections.

Abhari et al. (2020) deployed a growth prediction and containment of COVID-19 in Switzerland using a previously developed Al model (EnerPOL) and big data simulation platform and predicted there will be 720, 73,300, and 83,300 deaths, recoveries, and infections, respectively, between 22 February and 11 April 2020. This model also reckons for pre-intervention as well as post-intervention transmission rate, incubation period, and others.

It is quite reasonable to expect that, at the initial stage of an epidemic/pandemic outbreak, there will be a paucity of useful and reliable data, which makes forecasting a very challenging proposition. Therefore, Fong et al. (2020) proposed a methodology named Group of Optimized and Multisource Selection, abbreviated as GROOMS, to forecast the COVID-19 outbreak from a small dataset. This methodology is an ensemble of five different types of forecasting methods including classical time series forecasting and self-evolving polynomial neural networks (PNN). This study observed that PNN has relatively better performance and lowest error rate and further extended to PNN+cf with corrective feedback in the optimization process (Fong et al. 2020).

Zixin Hu et al. (2020) deployed a modified stacked auto-encoder Al model for real-time forecasting of COVID-19 and also estimated the size, lengths, and ending time of the pandemic. The data spanned from 11 January to 27 February 2020 and forecasted confirmed cases of COVID-19 from 20 January to 20 April 2020 in China. Further, dynamic patterns of the transmission of virus across the provinces/ cities were clustered, and the model also predicted that COVID-19 epidemics would be over by the middle of April (Hu et al. 2020), though retrospectively, it turned out to be an inaccurate proposition for the larger part of the globe. Therefore, it is warranted to be aware of the possible pitfalls of model development and reliance on their predictions before they have been validated by extensive research.

Al can be applied to track the coronavirus outbreak by requisite information from social media, websites, news reports, and other sources of data related to COVID-19 symptoms such as fever, breathing, and coughing as indicators, and use this information to predict where the disease is most likely to spread. For example, a company called BlueDot is using foreign-language news reports, animal and plant disease networks, and other official data sources for their Al model to predict, alert, and give advance warning to its clients about the next hotspots (Niller 2020). Thus, it can be argued that Al can cause a paradigm shift to the current COVID-19 or future global pandemic outbreaks by tracking and predicting the outbreak and curbing the spread.

Data Authentication: A Case for Blockchain Integration for Diverse AI Systems

Currently, there is no vaccine available with proven clinical efficacy against COVID- 19 (Cennimo 2020). Several clinical trials are ongoing and the World Health Organization (WHO) is playing a central role in reviewing the evidence generated by these trials. However, COVID-19 cases are expected to rise substantially in the coming days, but the supply of protective equipment, testing kits, healthcare staffs, etc., are limited (Livingston et al. 2020) and may lead to overwhelming of healthcare systems in many parts of the world as happened in Italy. To address the ongoing COVID-19 outbreak and accelerate the research with limited resources, we need open science, data sharing, and collaboration between researchers, academia, governments, official organizations, civil society, and the private sector to better monitor, understand, and accelerate COVID-19 research to mitigate this pandemic (United Nations SDSN Report 2020; White 2020). Even crowdsourcing can be a very effective, economical, and robust tool. Smith et al. (2015) discussed various forms of crowdsourcing and their advantages and studied the impact of crowdsourcing on economic resilience with the help of a case study.

One of the main problems with such open science and data sharing strategies is the availability of safe, secure, scalable, and verified data which will require collective efforts of clinicians, scientists, and researchers. Nevertheless, researchers are using blockchain technology in an attempt to mitigate some of these issues by distributed consensus algorithms which possess the following features: decentralized, transparent, immutable data along with autonomy, open source, and anonymity (Liang et al. 2016).

Shen et al. (2019) studied efficient management of patient medical records and proposed a mechanism for sharing medical records using the MedChain model. The model was developed on a decentralized platform, connecting various healthcare stakeholders utilizing the blockchain and peer-to-peer networks. It employed a session-based data sharing scheme which included data generation, session management, and key management. The data generation process consisted of collecting patient data through medical devices and integrating it into blockchain and directory services. During the session management process, a patient can concede access to their medical data to a requester through a session following different protocols. Key management applications in basic terms provided each participant with a software key case for storing received cryptographic keys. They have evaluated the performance and accuracy of this entire system based on two critical parameters, i.e., security and efficiency. Results indicated that the MedChain scheme was highly flexible, efficient, and secure compared to other existing models, without compromising on potential industrial scalability.

Other researchers like Gordon and Catalini (2018) have highlighted the role of blockchain in enabling patient-centric control of medical data over institution-centric control. They have identified five mechanisms consisting of digital access rules, data aggregation, data liquidity, patient identity, and data immutability through which blockchain can address various challenges. Daisuke et al. (Ichikawa et al. 2017) proposed a Hyperledger fabric blockchain platform with a unique way to collect data using smartphones and introduced a concept called the “mHealth system.” Stephen et al. (2018) identified the Ethereum platform as a potential candidate for managing healthcare data. Khezr et al. (2019) published a comprehensive review of the healthcare ecosystem, discussed some possible challenges of blockchain technology, and proposed an innovative delivery system called the internet of medical things (IoMT).

Currently, IBM have introduced “MiPasa” a global-scale control and communication system based on blockchain which manages COVID-19 outbreaks securely and democratically (Singh and Levi 2020). MiPasa utilizes blockchain technologies to gather reliable, quality data, address their inconsistencies, and make it easily accessible to technologists, data scientists, and public health officials. Omar et al. (2017) outlined a patient-centric healthcare data management system by using blockchain, which ensures privacy, integrity, accountability, and security of patients’ data with cryptographic functions. Yue et al. (2016) proposed an app architecture called “Healthcare Data Gateway” (HGD) for personalized healthcare, enabling the patient to access, monitor, and manage their own data easily and securely store on a private blockchain. Zhou et al. (2018) proposed a blockchain-based medical insurance storage system and developed it on the Ethereum platform for patients, emergency clinics, insurance agencies, and servers. The proposed blockchain-based framework has three layers, namely user layer, system management layer, and storage layer to secure privacy and autonomy of the data. Studies related to managing patient records using the blockchain platform have made tremendous progress in recent times (Agbo et al. 2019; Khatoon 2020; Khezr et al. 2019).

Estonia is an outstanding example of the application of blockchain technology in the healthcare system (Aaviksoo 2020). During an emergency (such as a pandemic), a doctor can access critical information of patients, such as blood types, allergies, recent treatments, ongoing medication, or pregnancy. They deployed KSI® blockchain technology for the system to ensure data integrity and to mitigate internal threats. Ministries/government agencies can also utilize these systems to measure health trends, track epidemics, and ensure a better distribution of healthcare resources. A few key takeaways from this initiative are as follows: (i) blockchain can store patients’ and clinical trial data for the development of an Al model to monitor, predict, diagnose, and treat COVID-19; (ii) blockchain can decentralize and democratize Al and machine learning model of COVID-19 to the rest of the world (Harris and Waggoner 2019); (iii) the blockchain can act like a central nervous system for an Al-linked network of sensors to predict important aspects of COVID-19 spread and outbreaks (Krittanawong et al. 2020); and (iv) with smart contract, blockchain can provide decentralized and autonomous pandemic organizations systems and Al can learn and adapt to changes over time to handle situations like COVID-19 pandemic without any delay and help regulatory agencies to take swift measures. Vulnerabilities of smart contracts can be detected by advanced techniques, further enhancing its security features (Tann et al. 2018; Gogineni et al. 2020).

< Prev   CONTENTS   Source   Next >