CHAPTER 2: LITERATURE REVIEW
2.4 What are the current distribution challenges for the SARS-CoV-2 treatment taken the
2.5.2 Federated Learning for Medical COVID-19 Applications
2.5.2.1 Literature survey for Federated Learning in Medical COVID-19 Applications 55
FL is a wide concept bringing the code to data rather than the data to the code and offers solutions to basic issues associated to data discretion, localisation and ownership. Federated fundamentals with ML have attracted considerable interest and have been used in many fields, especially in the medical sector. The frequency of their use has considerably increased after the outbreak of the SARS-COV-2 pandemic. Thus, patient data sharing and privacy have become main concerns.
Table 2.3 lists the eight studies that explored the use of FL for COVID-19 applications.
Table 2. 3 Literature Survey for Federated Learning in Medical COVID-19 Applications Author & Year Contribution Case Study Federated
Fundamental
Yang, Xu et al. (2020)
This study proposed a novel federated semi-supervised learning
technique (with or without
annotations) for effectively utilising available data in addressing
variability in both data and annotations.
Chest computed tomography
Federated and semi- supervised learning as traditional ML
Akhil Vaid and Suraj K Jaladanki et al.
(2021)
This study aimed to use federated learning technique to prevent local accumulation of raw clinical information across several institutions and forecast death in hospitalised COVID-19-positive
Electronic Health Records
Least absolute shrinkage and selection operator federated
Multilayer
perceptron federated as traditional ML
patients within 7 days.
Liwei Ouyang, Yong Yuan and Yumeng Cao et al. (2021)
The authors
proposed a
collaborative early warning style associated to blockchain and smart contracts for COVID-19, and the
goal was
crowdsourcing early warning
errands to
distributed stations in medical
institutions, people and other social strata.
Early warning for COVID-19 based on blockchain and smart contracts
The strategy combines the monitoring results of two screening methods, medical federation screening based on federated learning and social association screening based on the learning markets method to alert new cases as traditional ML
Wang, Xu, Ma, Talha, Al-Rakhami and Ghoneim et al. (2021)
This study proposed a 5G-enabled auxiliary diagnosis architecture based on federated learning for many institutions and central cloud cooperation to enable the sharing
of high
generalisation performance diagnosis models
5G-enabled federated learning auxiliary diagnosis of COVID-19
COVID-19 severity classification
Experiments were run on a central cloud and three edge cloud servers to ensure that the suggested architecture and model cognition technique were effective similar to traditional ML
Kumar, Khan, Zhang, Yang, Golilarz, Zakria, Ali, Shafiq
This study proposed a framework that collects a limited quantity of data from various sources (different hospitals) and uses
COVID-19 detection using CT imaging
Deep learning
and Wang et al.
(2020)
blockchain-based federated learning to train a global deep learning model
(Pang, Huang, Xie, Li and Cai et al. (2021)
This study proposed a framework that combines digital twins (DTs) of cities with federated learning to provide a
novel collaborative paradigm that enables numerous city DTs to share local
strategies and status quickly
City DTs to share local strategies and status quickly
Federated learning central server manages local updates from multiple
collaborators (city DTs), generating a worldwide
framework focused
on manifold
iterations at various city DT policies until the framework learns associations among different response stratagems and contamination trends, such as traditional ML (Zhang, Zhou, Lu,
Wang, Zhu, Sun, Wang, Lo, Wang, et la., 2021)
This study proposes a novel federated learning
approach based on dynamic fusion for medical diagnostic imagery analysis to diagnose
Coronavirus comorbidities.
COVID-19 Detection using Dynamic Fusion- based Federated Learning
DYNAMIC FUSION-BASED FEDERATED
LEARNING as
traditional ML
Feki, Ammar,
Kessentini, and Muhammad et al.
(2021)
This study proposed a federated learning system that allows several medical organisations to
apply deep
knowledge to scrutinise the virus from chest X-ray
Screening COVID-19 from chest X-ray images
Federated learning of a deep CNN model
imagery while maintaining
confidentiality.
As shown in Table 2.3, eight articles focused on the use of FL in medical applications, such as COVID-19 detection using CT imaging, early warning for coronavirus based on blockchain and shrewd comparison and local strategy and status share of city DTs. Federated fundamentals well suited and widely used with ML models. Six of eight studies contributed to traditional ML, and the remaining studies contributed to deep learning that overcome the first obstacle of privacy (data sharing issue). However, no research explored prioritisation challenges. In other words, the second challenge cannot be addressed, which requires a precise decision-making approach to resolve the three prioritisation issues outlined. This situation is considered a research gap.
2.5.3 Point of view for Multicriteria decision making & Treatment Distribution
Apart from analysis, MCDM is regarded as a solution that assists experts in organising and solving any prioritisation issue (M. A. Alsalem et al., 2021). MCDM is defined as a decision theory extension covering any decision with multiple objectives and is a technique for evaluating options according to several contradictory criteria and for merging them into a single inclusive evaluation (M. A. Alsalem et al., 2021). Therefore, MCDM provides great benefits as a decision science support technique and has been used in the context of SARS-COV-2 along with a variety of applications. Identifying the prioritisation issue (second challenge) of SARS-COV-2 treatment recipients as an MCDM issue has been discussed in the literature. T. J. Mohammed et al. (2021) proposed an intelligent framework based on the MCDM context and successfully addressed patients’ prioritisation issue in distribution hospital networks and issues in the transfusion of
efficient CP from donors to the most critical SARS-COV-2 patients as an early treatment (i.e. pre- vaccination stage). However, this framework relied on sharing donors’ and patients’ data in all hospitals (distribution hospitals) and combined them in a decision matrix for donors and in another decision matrix for SARS-COV-2 patients for prioritisation and matching. The privacy and sensitivity of patients’ and donors’ openly used data were ignored.
To date, no study has integrated the federated fundamental with MCDM techniques for SARS- COV-2 treatment. To bridge this gap, integration is essential to characterise the supply of anti- SARS-CoV-2 mAbs in distribution hospitals. Therefore, the formulation of a new federated fundamental concept called ‘Federated-Decision Making Distributor (FDMD)’ is necessary to overcoming challenges towards ensuring the privacy of health SARS-COV-2 data and prioritising anti-SARS-CoV-2 mAb recipients in distribution hospitals.