Conceptualization, K.S.; methodology, K.S.; software, R.K., A. and D.S. and K.S.; validation, K.S. and Y.-C.H.; formal analysis, R.K., A. and D.S.; investigation, R.K., A. and D.S.; resources, K.S. and Y.-C.H.; data curation, R.K., A. and D.S.; writing—original draft preparation, R.K., A., D.S. and K.S.; writing—review and editing, R.K., A., D.S., K.S. and Y.-C.H.; visualization, R.K., A., D.S. and K.S.; supervision, K.S. and Y.-C.H.; project administration, Y.-C.H.; funding acquisition, Y.-C.H. All authors have read and agreed to the published version of the manuscript.

Figure 1. Artificial Intelligence—Nomenclature.

Figure 1. Artificial Intelligence—Nomenclature.

Figure 2. AI and blockchain relationship.

Figure 2. AI and blockchain relationship.

Figure 3. PRISMA flow diagram for the selection process of the research articles used in this review.

Figure 3. PRISMA flow diagram for the selection process of the research articles used in this review.

Figure 4. Time graph—number and year of publications studied in this review.

Figure 4. Time graph—number and year of publications studied in this review.

Figure 5. Block diagram representing the structure of this review.

Figure 5. Block diagram representing the structure of this review.

Figure 6. Blockchain layered structure.

Figure 6. Blockchain layered structure.

Figure 7. Blocks are linked together in a blockchain using a cryptographic hash. x is an arbitary block, x + 1 is a suceeding block and so on.

Figure 7. Blocks are linked together in a blockchain using a cryptographic hash. x is an arbitary block, x + 1 is a suceeding block and so on.

Figure 8. Utilities of AI in public health.

Figure 8. Utilities of AI in public health.

Figure 9. Concept of an artificial neural network for healthcare.

Figure 9. Concept of an artificial neural network for healthcare.

Figure 10. Algorithm implementing coupled K-means Clustering and Naive Bayes Algorithm.

Figure 10. Algorithm implementing coupled K-means Clustering and Naive Bayes Algorithm.

Figure 11. Theoretical representation of a Deep Recurrent Neural Network.

Figure 11. Theoretical representation of a Deep Recurrent Neural Network.

Figure 12. Theoretical representation of a Deep Belief Network.

Figure 12. Theoretical representation of a Deep Belief Network.

Figure 13. Basic structure of a Deep Convolutional Neural Network.

Figure 13. Basic structure of a Deep Convolutional Neural Network.

Figure 14. Illustration of a system where a provider adds an EHR for new patients using blockchain.

Figure 14. Illustration of a system where a provider adds an EHR for new patients using blockchain.

Figure 15. Data processing model of EHRs for remote monitoring.

Figure 15. Data processing model of EHRs for remote monitoring.

Figure 16. The basic structure of MIStore. Arrows represent the flow of data while the numbers represent the sequence of the processes.

Figure 16. The basic structure of MIStore. Arrows represent the flow of data while the numbers represent the sequence of the processes.

Figure 17. Open challenges in using AI-powered Blockchain for public health.

Figure 17. Open challenges in using AI-powered Blockchain for public health.

Figure 18. Future research directions for AI-powered Blockchain in public health.

Figure 18. Future research directions for AI-powered Blockchain in public health.

Table 1. List of abbreviations used along with their full form.

Table 1. List of abbreviations used along with their full form.

AcronymDefinitionAIArtificial IntelligenceMLMachine LearningDLDeep LearningCNNConvolutional Neural NetworkANNArtificial Neural NetworkNLPNatural Language ProcessingDRLDeep Reinforcement LearningRNNRecurrent Neural NetworkGNNGraph Neural NetworkKNNK-Nearest NeighborDGMDeep Generative ModelDRLDeep Reinforcement LearningPIIPersonally Identifiable InformationEHRElectronic Health RecordLSTMLong-Short Term MemoryMBRModel-Based Reasoning

Table 2. Comparison with other similar review articles.

Table 2. Comparison with other similar review articles.

Reference and YearNumber of ArticlesPeriodOne-Phrase SummaryBlockchainAI-Powered TechniquesOpen ChallengesFuture DirectionsMachine LearningDeep LearningThis Review1202012–2022A systematic review of AI and Blockchain powered techniques in healthcare✓✓✓✓✓[6], 20212081998–2021Discusses security and privacy implications of applying ML/DL techniques in healthcare✗✓✓✓✓[7], 20211502002–2021Ethics of implementing ML in healthcare✗✓✗✓✗[8], 2021682016–2020Review on the usage of data from health devices in ML-based healthcare systems.✗✓✓✗✗[9], 20211592008–2021Integrating ML models into IoT-based healthcare devices✗✓✓✓✗[10], 20211582014–2021Review of smart healthcare systems including usage of wearable devices and smartphones for monitoring.✗✓✓✓✓[11], 2019722014–2019ML adoption for Blockchain-based systems to make them more resistant to attacks.✓✓✓✓✗

Table 3. A summary of works on Machine Learning-Enabled Blockchain Technologies for Public Health.

Table 3. A summary of works on Machine Learning-Enabled Blockchain Technologies for Public Health.

Ref.Healthcare ApplicationSecurity ChallengesDatasetMachine Learning Approaches UsedBlockchain Types UsedBlockchain PlatformKey ContributionLimitations[53]Personal Health Record (PHR) verificationEHRs are not used much because of privacy and security concernsSample medical data of patients are used.Neural NetworksBlockchain using Hyper POR algorithmEthereumVerification of medical information with accurate image data extraction and blockchain based on PHR.Lack of legal and institutional mechanisms that can support activation of blockchain.[54]Storing EHRs, clinical trial reports and creating machine learning models based on this dataInsecure storage of highly private and critical data--Depending upon the needs, various unsupervised and supervised algorithms can be implementedStandard tamper-proof append only ledger, much like the cryptocurrency Bitcoin’s Ledger---Developing a secure and trusted data management platformEfficiency of the models will depend on the quality of data stored in the blockchain and the degree of tamper resilience will depend on how well the blockchain has been implemented.[55]Quick and efficient storage, retrieval, and sharing of electronic medical records and smart model for updated report generationInsecure transfer of dataThe data uploaded by the patients through the network is used as the dataset for the training purposes.Natural language processing, Convolutional Neural Networks, Image Processing, and Encoding-Decoding.No particular type has been mentioned in the text.Ethereum-Based BlockchainSmart electronic medical report handling and UsageBugged Ethereum contracts can render the system highly inefficient.[56]Developing a voting-based smart diagnosis system where data can be stored and shared on a blockchain among various health centres.--The data present in the blockchain will serve as the training data for the models present with each healthcare centreArtificial Neural Networks, Support vector machines, Decision trees, Random Forests, AdaBoost, and Bayesian NetworksNo particular type has been mentioned in the textNo particular type has been mentioned in the textSmart voting-based disease diagnosis systemThis system will be as efficient as the quality of data present on the blockchain.[57]Handling patient data, health records, AI-powered Diagnosis, optimized testing, Radiological and Psychological deductions, smart drug deliveryPrivacy and Anonymity, Usability of data, and consent of patients and organizations.Available data on the blockchain serves as the datasetsRandom Forest Classifiers and Regression, Deep learning algorithms such as Convolutional Networks and Recurrent Neural Networks.No particular type has been mentioned in the textNo particular type has been mentioned in the textDifferent spheres of Medical Diagnostics and data management that can benefit from the combination of machine learning and Blockchain technologies.---[58]Using machine learning and blockchain technology for Cancer care--Combination of data present in different institutions and organizations connected by blockchain along with environment-based factors and regional healthcare service assessment of the survivorsNo particular algorithms have been named in this workNo particular type has been mentioned in the textNo particular type has been mentioned in the textConceptualized a system that can detect cancer recurrence and prognosis using machine learningThis system will be as efficient as the quality of data present on the blockchain[59]Sharing personal continuous-dynamic health data.Centralised data storage methods hinder sharing and has single point of attack--Conceptual usage of advanced machine learning algorithms.Public blockchain typeEthereum, Hyperledger FabricData sharing and transaction validation along with quality validation module using machine learning.Security issues once the data are transferred to the customer are not discussed.

Table 4. A summary of works on deep learning-enabled Blockchain Technologies for Public Health.

Table 4. A summary of works on deep learning-enabled Blockchain Technologies for Public Health.

Ref.Healthcare ApplicationSecurity ChallengesDatasetDeep Learning Model UsedBlockchain Type UsedBlockchain PlatformKey ContributionLimitations[61]Blockchain-deep learning environment for analyzing EHRs.EHR needs to be secured against unauthorized access and cyber-attacks.EHR data from some hospitals are used after permission.RNN-Long Short-Term Memory (RNN-LSTM).Permissioned blockchain network/private blockchain.Hyperledger.Store EHR of patients using Hyperledger and analyze blocks using RNN-LSTM and RNN-GRU.High maintenance coset compared to traditional models.[59]Detection of myopiaData transfer and sharing between collaborators for medical studies.Retinal photographs from Singapore, China, Taiwan, and India.A new model developed to detect myopia.Permissioned blockchain network/private blockchain.Hyperledger.Detecting internal diseases such as Myopia.The severity of the problem is not identified.[69]Secure way of sharing EHRs amongst healthcare users.Threat to availability and integrity of the EHRs. Collusion attacks are also possible in the given scenario.EHR records from different hospitals.A deep learning model consisting of 4 layers is suggested.In the suggested blockchain model, n lattices-based cryptography has been used.Post-quantum blockchain networks are used.A blockchain-based deep learning as-a-service model is suggested which ensures secure and accurate sharing of EHRs.Scalability and complexity of the proposed model.[70]To carry out accurate DNA diagnosis of Malaria and blockchain technology is used for secure management of the dataset.Secure data connectivity between different geo-locations in rural areas is quite difficult and the management of the dataset is also essential.Dataset was taken from field tests in rural Uganda.----Hyperledger Composer.----[71]Blockchain convolution neural networks and audio-video emotion patterns to detect healthcare emergencies in nearby areas.Providing security to patients’ reports and other medical services is quite essential and there can be a major threat of external parties accessing the patient’s details.Audio-visual patterns will help in training the model of emotional recognition.CNN deep learning techniques are used.Blockchain convolution neural networks.Ethereum ecosystem has been used.Contributed to the field of emotional recognitions using deep learning and it provides a layer of security as well using smart contracts.Not included recurrent network methods which can help to predict results for an existing or past patient.[72]Suggested a secure model with the application of blockchain framework Exonum and deep learning algorithms to manage control over personal data in medical records.The data possessed by patients in form of their medical records hold great value for predictive analysis and securing that data, which is limiting access to the patient and doctor, it possesses a major challenge.The dataset of some patients has been taken with their consent to apply different deep learning and blockchain algorithms.GAN-generative adversarial networks.Exonum framework blockchain.Exonum.Suggested a system that can reduce the difficulty of carrying out biomedical research.Limited to text-form of data and other forms have not been considered in the database.[73]The proposed model is formed to identify feature-extracted data from the existing data. Deep Neural networks and blockchain smart contracts have been applied to the model.EHRs have a centralized database and that is a major security issue since the control belongs to a single person.Greater Noida COVID-19 dataset is used.The DNN model is trained for feature extraction. The major diseases in a particular area.Blockchain with a token-based approach is used.Ethereum virtual machine.The bulk data are reduced to data size which can simply be predictive analysis.More advanced versions of blockchain could have been used.